Bug#831077: Preparing for upload
Hi, to make some progress here, I would like to upload a new version of casacore based on the current git repository by the end of the week. This would close this bug. I would also add myself into the "Uploaders" list (and mark it as a regular upload). Are there any objections? Gijs? Benda? Cheers Ole
Bug#798818: Reopen bug for updating kernel for mipsel buildds
Hi Aurelien, On 27.09.2016 13:31, Aurelien Jarno wrote: > We have identified an hardware FPU bug on the Loongson 3 build daemons, > a dozen of packages are known to be affected so far, it seems your > package is an additional one. > [...] > As explained above, this won't fix your issue. That said I have > configured those build daemons to not build casacore anymore, and I have > given back casacore. This should fix your issue. Thanks a lot for the explanation and adding casacore to the blacklist. This helps a lot to get through the portability issues of casacore :-) Cheers, and greetings to Lyon Ole
Bug#831175: Bug is not reproducible anymore
Control: tags -1 + unreproducible I cannot reproduce this anymore; also the just-uploaded new version (with minor changes not affecting the build) builds fine. The cause of the FTBFS is probably not in the package itself, but in the "casacore" dependency, which was not built with gcc-6 at the time of the bug report. I fixed this (and also some other problems) in the casacore package. If there are no more build problems on x86_64, I would close this bug in the next days. Cheers Ole
Bug#798818: Reopen bug for updating kernel for mipsel buildds
unarchive 798818 reopen 798818 thanks Sorry for disturbing you again: When building on mipsel-aql-02.debian.org, the kernel is still Kernel: Linux 3.16.0-4-loongson-3 mipsel (mips64) which is far outdated, and causes an FTBFS: https://buildd.debian.org/status/fetch.php?pkg=casacore=mipsel=2.1.0-3=1474903782 For comparison, the previous one succeeds: https://buildd.debian.org/status/fetch.php?pkg=casacore=mipsel=2.1.0-2=1474385578 Similarly, I get a failure on mipsel-aql-02.debian.org (this time for mips64). Could you please update *all* mips buildds to a 4.7 kernel? Otherwise there is always trouble with the FPU.
Bug#830254: Failures due to kernel bugs in MIPS
The three tests * tClassicalStatistics * tHingesFencesStatistics * tLCEllipsoid FTBFS on MIPS platforms (mips, mipsel, mips64) when they are build on linux kernel version 3.16.0, but succeed on kernel version 4.7. Since I previously had floating point problems on MIPS with older kernels -- see f.e. https://bugs.debian.org/811368 https://bugs.debian.org/781892 They *should* be all updated; see https://bugs.debian.org/798818 However, mipsel-aql-* machines are still on kernel 3.16...
Bug#839348: astroml: FTBFS: dh_auto_test: pybuild --test --test-nose -i python{version} -p 2.7 returned exit code 13
Control: reassign -1 python-sklearn 0.18-1 Control: retitle -1 python-sklearn: wrong import of "parallel" Control: affects -1 src:astroml This is a bug in scikit-learn, not in astroml: $ python Python 2.7.12+ (default, Sep 1 2016, 20:27:38) [GCC 6.2.0 20160822] on linux2 Type "help", "copyright", "credits" or "license" for more information. >>> import sklearn.mixture Traceback (most recent call last): File "", line 1, in File "/usr/lib/python2.7/dist-packages/sklearn/mixture/__init__.py", line 5, in from .gmm import sample_gaussian, log_multivariate_normal_density File "/usr/lib/python2.7/dist-packages/sklearn/mixture/gmm.py", line 27, in from .. import cluster File "/usr/lib/python2.7/dist-packages/sklearn/cluster/__init__.py", line 6, in from .spectral import spectral_clustering, SpectralClustering File "/usr/lib/python2.7/dist-packages/sklearn/cluster/spectral.py", line 16, in from ..metrics.pairwise import pairwise_kernels File "/usr/lib/python2.7/dist-packages/sklearn/metrics/__init__.py", line 33, in from . import cluster File "/usr/lib/python2.7/dist-packages/sklearn/metrics/cluster/__init__.py", line 20, in from .unsupervised import silhouette_samples File "/usr/lib/python2.7/dist-packages/sklearn/metrics/cluster/unsupervised.py", line 13, in from ..pairwise import pairwise_distances File "/usr/lib/python2.7/dist-packages/sklearn/metrics/pairwise.py", line 27, in from ..externals.joblib import parallel ImportError: cannot import name parallel It is a regression; 0.17.1-2 (in testing) works fine. Best regards Ole
Bug#843668: ITP: drizzle -- Dithered image combination for Python
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-Cc: debian-as...@lists.debian.org, debian-de...@lists.debian.org * Package name: drizzle Version : 1.1 Upstream Author : Bernie Simon <bsi...@stsci.edu> * URL : http://spacetelescope.github.io/drizzle/ * License : BSD-3-Clause Programming Lang: Python Description : Dithered image combination for Python The drizzle library is a Python package for combining dithered images into a single image. This library is derived from code used in drizzlepac. Like drizzlepac, most of the code is implemented in the C language. The biggest change from drizzlepac is that this code passes an array that maps the input to output image into the C code, while the drizzlepac code computes the mapping by using a Python callback. Switching to using an array allowed the code to be greatly simplified. The package is an upcoming astropy affiliated package and will be maintained by the Debian Astro Team. A git repository is created at https://anonscm.debian.org/git/debian-astro/packages/drizzle.git Best regards Ole
Bug#843667: ITP: drizzle -- Dithered image combination for Python
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> * Package name: drizzle Version : 1.1 Upstream Author : Bernie Simon <bsi...@stsci.edu> * URL : http://spacetelescope.github.io/drizzle/ * License : BSD-3-Clause Programming Lang: Python Description : Dithered image combination for Python The drizzle library is a Python package for combining dithered images into a single image. This library is derived from code used in drizzlepac. Like drizzlepac, most of the code is implemented in the C language. The biggest change from drizzlepac is that this code passes an array that maps the input to output image into the C code, while the drizzlepac code computes the mapping by using a Python callback. Switching to using an array allowed the code to be greatly simplified. The package is an upcoming astropy affiliated package and will be maintained by the Debian Astro Team. A git repository is created at https://anonscm.debian.org/git/debian-astro/packages/drizzle.git Best regards Ole
Bug#843709: Purify missed the cfitsio5 transition
Package: release.debian.org User: release.debian@packages.debian.org Usertags: binnmu Severity: normal Hi, The "purify" package was in NEW while the cfitsio transition was going on, so it missed it. Please do an binNMU for the package now. nmu purify_2.0.0-1 . amd64 . -m 'Rebuild against libcfitsio5' Thanks Ole
Bug#840094: blends-dev: Does not recognize multiline dependencies
Hi Andreas, On 09.11.2016 12:47, Andreas Tille wrote: > In other words: Once it was defined as syntax for these control files > that newlines need to be escaped. I do not like it and as I said this > is fixed in the long-term pending rewrite. However, the bug is not > serious but at best wishlist. Would you follow this arguing? > Not really. My point here is that this happens really unexpected, and since blend-gen-control doesn't complain about the then wrong format, one silently gets wrong dependencies. At least I did in the first versions of debian-astro (<0.5). We have a clear definition of how these files should look like, namely RFC822, and this also defines continuation lines. Look at https://blends.debian.org/blends/ch08.html#edittasksfiles - it is blends-gen-control that isn't conform to that. I would think that there is also a quick fix for it -- the tool already handles continuation lines for the tasks description, so one could probably just take that. I have no glue of all the Perl $@^!~ special chars, but wouldn't do it something like the attached patch (after removing the obvious errors from it)? Or something else just adopted from lines 556-562 of blends-gen-control? Best regards Ole diff --git a/devtools/blend-gen-control b/devtools/blend-gen-control index 1aba552..cde3237 100755 --- a/devtools/blend-gen-control +++ b/devtools/blend-gen-control @@ -566,9 +566,14 @@ sub load_task { my $header; for $header (qw(Depends Recommends Suggests)) { if (m/^$header:\s+(.+)$/ && $1 !~ /^\s*$/) { + my $pkgs = $1; + while () { + last if (m/^\S+/ || m/^\s*$/); + $pkgs .= $_; + } $taskinfo{$curpkg}{$header} = () if (! exists $taskinfo{$curpkg}{$header}); -my ($pkglist, $missinglist) = process_pkglist($1); +my ($pkglist, $missinglist) = process_pkglist($pkgs); push(@{$taskinfo{$curpkg}{$header}}, @{$pkglist}); $haspackages += $#{$taskinfo{$curpkg}{$header}} + 1;
Bug#840094: blends-dev: Does not recognize multiline dependencies
Hi Andreas, Petter and all, On 10.11.2016 21:07, Andreas Tille wrote: > So I confirm that the first problem we detected is solved but there is > another one breaking Debian Edu. I have again no suspicion why the '\' > sign is not elimiunated from the list only in those few cases. I can also just "poke in the fog" here. I thought that the multilines were already eaten up by lines 537-544, and should not appear again here. However, they do, as my "print" debugger shows. Perlists, please help!!! The pragmatic solution here is to just remove the backslashes from the end of line. I can commit a patch that does this. However, debian-edu keeps to be broken. Reason is that the tasks contain lines like (development) Depends: fp-compiler, [...multiline... ], fp-units-fv Depends: lazarus so, with two "Depends" in one section. Or (electronics): Depends: qucs, gpsim, oregano Recommends: education-menus, xoscope Suggests: kicad, kicad-doc-en, kicad-doc-de, kicad-doc-es, kicad-doc-fr Suggests: electric, pcb, xcircuit, freehdl, gtkwave Responsible: ? NeedConfig: no two "Suggests". This does not work anymore (no idea why), but is also IMO forbidden by RFC834. I have no idea why this works without RFC834 continuation, but not with them. On the other hand, we *win* one more dependency: in "common", the "firmware-ralink" dependency line was *not* preceded one with a backslash. This shows that the backslash is just a bad idea for continuation lines. --> debian-edu tasks are just broken. They don't follow any rule, and depending from the parser one will get always different results. Maybe we should fix that? remove all backslash continuation lines and duplicated keywords from the tasks files? Best regards Ole
Bug#840094: blends-dev: Does not recognize multiline dependencies
Hi all, On 11.11.2016 08:07, Andreas Tille wrote: > On Thu, Nov 10, 2016 at 10:38:32PM +0100, Ole Streicher wrote: >> --> debian-edu tasks are just broken. They don't follow any rule, and >> depending from the parser one will get always different results. Maybe >> we should fix that? remove all backslash continuation lines and >> duplicated keywords from the tasks files? > > I think it should be sufficient to add line breaks where needed. If we touch them anyway, we could make them fully RFC compliant at this time as well. Since I am currently stuck in the S-Bahn (Polizeieinsatz), I will do that for debian-edu and push. Cellphone internet is a nice thing sometimes... :-) @Petter, please review and change/revert if you disagree. Cheers Ole
Bug#840094: blends-dev: Does not recognize multiline dependencies
Hi Andreas, On 10.11.2016 13:48, Andreas Tille wrote: > Hi Petter, > > On Thu, Nov 10, 2016 at 12:39:07PM +0100, Petter Reinholdtsen wrote: >> Control: tags -1 + pending >> >> [Andreas Tille] Should I commit it? >>> Yes please. Ole, you reported problems with your patch. Could you >>> please be more verbose about this problem - at best based on Petter's >>> commit to make sure we are all working on the same code? >> It is now commited. Please give it some more testing before uploading. >> Preferably also with the debian-edu git repo, where both tasks and >> control file is kept in git, allowing any changes to be easily >> discovered. > I need to admit two things: Even in Debian Med we went into the trap > criticised by Ole. In bio-cloud which is not maintained by me we were > also loosing entries. The second thing I need to admit that there are > in fact syntax errors resulting from the patch. When using the new > blends-dev package a > > gbp clone ssh://git.debian.org/git/blends/projects/med.git > cd med > make dist > grep ^, debian/control > > shows the problem. It leads to something like > > Suggests: bagpipe, > cufflinks, > embassy-phylip, > gmap, > r-cran-vegan > , > r-other-mott-happy.hbrem > , > r-other-valdar-bagphenotype.library, > soapdenovo2 > , > sra-toolkit > , > staden-io-lib-utils > , This is what I meant, and it is fixed by my last commit. Please try again ;-) Cheers Ole
Bug#840094: blends-dev: Does not recognize multiline dependencies
Hi Andreas & all, On 09.11.2016 15:35, Andreas Tille wrote: >> We have a clear definition of how these files should look like, namely >> RFC822, and this also defines continuation lines. > > Unfortunately in this specific feature tasks files are not RFC822 > compliant, which sucks, yes. Its even not documented (I just checked > since I intended to document it at some point in time but can't find it > :-( ) If one of our main tools is not compliant with our documented specifications, and this may cause incomplete metapackages (which are one central part of the blend), then I would still rate this as an RC bug, independently of whether it is easy to fix or not. >> I would think that there is also a quick fix for it -- the tool already >> handles continuation lines for the tasks description, so one could >> probably just take that. I have no glue of all the Perl $@^!~ special >> chars, but wouldn't do it something like the attached patch (after >> removing the obvious errors from it)? >> >> Or something else just adopted from lines 556-562 of blends-gen-control? > > While I fully agree that we should fix this I'm not fully convinced how > to sensibly proceed here. The problematic thing is that we are quite > short before a release and if we might break metapackage creation in > some way we might get in trouble. I'm no Perl programmer myself (even > if I think your patch looks sensible) and so IMHO staying conservative > and add some line ending escapes could be the less invasive change. I checked my patch, and it does *not* work correctly, it will produce syntax errors in the debian/control file, if RFC822 continuation lines are used. For tasks that have all in one line, or that have metapackages, everything seems to be OK, however. > If you (and Bas and other readers here) think we should fix the issue > right now I'm fine if you apply the patch below and we should seriously > test the metapackage creation of each Blend *before* 2016-12-05. > > What do you think? I am ready to test and also to fix; however my know-how ends here. I don't know what is wrong with the fix. Just wondering, and starting to really get worried: None of the debian-blends maintainers has enough Perl knowledge to fix this? If we all do not know Perl, why do we use that language in one of our central tools? That sounds to me even more RC than the bug itself... Cheers Ole
Bug#840094: blends-dev: Does not recognize multiline dependencies
Hi Andreas and Bas, On 10.11.2016 08:45, Andreas Tille wrote: > On Wed, Nov 09, 2016 at 06:27:13PM +0100, Sebastiaan Couwenberg wrote: >> On 11/09/2016 03:35 PM, Andreas Tille wrote: >>> If you (and Bas and other readers here) think we should fix the issue >>> right now I'm fine if you apply the patch below and we should seriously >>> test the metapackage creation of each Blend *before* 2016-12-05. >>> >>> What do you think? >> >> I think supporting the deb822 format should be a Blends release goal for >> buster, > > I fully agree. I admit I did way less for blends-dev than I intended to > do but other tasks that felt more urgent occupied all my Debian time. > >> it's indeed to late in the stretch dev cycle in my opinion. > > That would mean to lower the severity of #840094. Ole, are you OK with > this? No, I am not. The tool gives wrong results, and it does so silently. If I would not have discovered this by chance, the Debian Astro tasks packages were still wrong and would have stayed so in Stretch. I don't know about other blends, whether they ever checked the metapackages -- others still may have missing tasks as well, just because their maintainers read the docs and trust the debian-blends package. And you can't just change the specs in the last minute. And, if the package is in Stretch, people will start their new blend with exactly this package, again running into this trouble. I would really propose to fix that before Stretch. It just can't be that difficult. Nobody knows a Perl specialist who can have a look? (Oh, and can we have Python for the Next Generation blends-dev? This is at least what I understand, and there is already a parser for rfc822) Best regards Ole
Bug#840094: blends-dev: Does not recognize multiline dependencies
Hi Andreas and Bas, On 10.11.2016 08:45, Andreas Tille wrote: > On Wed, Nov 09, 2016 at 06:27:13PM +0100, Sebastiaan Couwenberg wrote: >> On 11/09/2016 03:35 PM, Andreas Tille wrote: >>> If you (and Bas and other readers here) think we should fix the issue >>> right now I'm fine if you apply the patch below and we should seriously >>> test the metapackage creation of each Blend *before* 2016-12-05. >>> >>> What do you think? >> >> I think supporting the deb822 format should be a Blends release goal for >> buster, > > I fully agree. I admit I did way less for blends-dev than I intended to > do but other tasks that felt more urgent occupied all my Debian time. > >> it's indeed to late in the stretch dev cycle in my opinion. > > That would mean to lower the severity of #840094. Ole, are you OK with > this? No, I am not. The tool gives wrong results, and it does so silently. If I would not have discovered this by chance, the Debian Astro tasks packages were still wrong and would have stayed so in Stretch. I don't know about other blends, whether they ever checked the metapackages -- others still may have missing tasks as well, just because their maintainers read the docs and trust the debian-blends package. And you can't just change the specs in the last minute. And, if the package is in Stretch, people will start their new blend with exactly this package, again running into this trouble. I would really propose to fix that before Stretch. It just can't be that difficult. Nobody has a Perl specialist who can have a look? (Oh, and can we have Python for the Next Generation blends-dev? This is at least what I understand, and there is already a parser for rfc822) Best regards Ole
Bug#845427: python-pywcs: ROM; dead upstream; superceded by python-astropy
Package: ftp.debian.org Severity: normal X-Debbugs-CC: debian-as...@lists.debian.org Dear ftpmasters, please remove python-pywcs. It is not maintained upstream anymore since years and completely superceded by python-astropy. Without major efforts, it is unusable because of incompatibilities to its dependencies. CI tests fail. People should just move to astropy and replace import pywcs with import astropy.wcs as pywcs There are no reverse dependencies. Best regards Ole
Bug#844087: Please consider removing pyfits
Hi Aurelien, On 22.11.2016 22:12, Aurelien Jarno wrote: > On 2016-11-12 11:40, Ole Streicher wrote: >> is however valid for pyfits, which also will not see any upstream love >> anymore. > Note that pyfits has seen an upload in 2016, so it got upstream love > recently, and there is a newer version than the one from Jessie in the > archive. That said it was clearly the last version. Sure, however the major reason for my proposal is that upstream itself expressed that one should switch. And he thinks about releasing a really-last version that just imports astropy.io.fits. >> IMO the release of Stretch is a good opportunity to force the users into >> the use of astropy, especially as it is as easy as replacing >> >> import pyfits >> >> by >> >> import astropy.io.fits as pyfits > That's not fully correct. That's true for simple programs, that said > astropy.io.fits has removed or deprecated some functions, so it doesn't > work in the most complex cases. Do you have an example here? From my experience, already the recent versions have some incompatibilities. These incompatibilites lead to a non-working pywcs, and this is the occasion for me to think about the pywcs removal. For old code, you just can't use pyfits anymore; header.update(key, value=v, comment=c) will not work. > My plan has always been to remove pyfits from the archive after the > release of Stretch. I don't see the point of doing that earlier. I am afraid that the user will migrate to the last pyfits release instead of astropy, even if the effort is the same. And stay in the dead end. Best regards Ole
Bug#845332: ITP: casacore-data-sources -- Table of ICRF reference source coordinates for casacore
Hi Jonathan, On 22.11.2016 20:12, Jonathan Quick wrote: > On Tue, November 22, 2016 5:22 pm, Ole Streicher wrote: >> It is still undecided if we take ICRF1 (1998; 608 sources) or >> ICRF2 (2007; 3414 sources). > > ICRF-2 supersedes ICRF-1 with more accurate positions. Currently the next > realisation (ICRF-3) is under preparation for release in 2017. I know; however casacore in the moment provides a table with ICRF1; see f.e. ftp://ftp.astron.nl/outgoing/Measures/WSRT_Measures.ztar If there is no reason to stay with ICRS1, I would use ICRF2; I will however see if some arguments pop up upstream: https://github.com/casacore/casacore/issues/539 Cheers Ole
Bug#845332: ITP: casacore-data-sources -- Table of ICRF reference source coordinates for casacore
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org * Package name: casacore-data-sources Version : 1 or 2 * URL : https://www.iers.org/IERS/EN/DataProducts/ICRF/icrf.html * License : Public domain Description : Table of ICRF reference source coordinates for casacore This package contains a table with the sources that realize the and the International Celestial Reference Frame (ICRF), as a table for the use with casacore. The ICRF is now the standard reference frame used to define the positions of the planets (including the Earth) and other astronomical objects. The package is a part of the effort taken by Benda Xu and me within the Debian Astro team to split up the original "casacore-data" package (RFP #761146) by data source into smaller packages which can be maintained independently. It is still undecided if we take ICRF1 (1998; 608 sources) or ICRF2 (2007; 3414 sources). The package will be maintained on our git repository under https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-sources.git Best regards Ole
Bug#846142: ITP: casacore-data-lines -- Table of spectral lines for casacore
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org * Package name: casacore-data-lines Version : 1.0 * URL : https://github.com/casacore/lines-table * License : CC0 Description : Table of spectral lines for casacore This package contains a table with the spectral line frequencies for the use with casacore. The package is a part of the effort taken by Benda Xu and me within the Debian Astro team to split up the original "casacore-data" package (RFP #761146) by data source into smaller packages which can be maintained independently. The package will be maintained on our git repository under https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-lines.git Best regards Ole
Bug#846002: blends-tasks must be priority:standard and not make a mess out of tasksel menu
Control: reassign -1 blends-tasks Control: tags -1 moreinfo Hi Holger, thank you for your bug report to the "blends" package. I would, however question a few things here and also ask for a little bit more information: The "blends-tasks" package was created as a result of working on bug #758116, titled "tasksel: Allow to select Blends selection during installation - just 'DE', 'Web server', 'Mail server' is NOT enough". This bug was created more than two years ago, and nowhere in the bug it was questioned that the blends should be selectable during the installation process. This, however, *requires* to have the information about the blends available during installation, and this makes the package that provides this information "important". Therefore, it is not a policy violation, which in turn removes your argument to make this bug "serious". It also does not "completely break the UI of the installer" -- the selection is in no mean different from the desktop environment selection. I would therefore propose to lower the severity of the bug. Also, I would ask you to do an NMU until the discussion has settled down. This would be an abuse of the NMU. NMUs are meant to deal with unresponsive maintainers, and you did not show any evidence that the blends maintainers are not responsive with respect to this problem. Doing NMUs during a discussion is quite offending. I also don't see a reason to hurry with implementing an unsettled decision: the blends selection is there since almost 8 months now without any significant change or discussion for ~6months. What makes the issue now so urgent that you try to push this within four days? Why didn't you do this half a year ago? We implemented the current solution at that time (and you *knew* that we did) exactly to have some buffer for discussion about critics. Why didn't you use that, but start now when it is quite late? The next point is that you base your critics not on some experience with the current installer but on an outdated, half-a-year-old screenshot. Since then, several improvements were done, both in the appearance in the installer, and in the selection of which blends are there. For the first, see the discussion here: https://lists.debian.org/debian-blends/2016/07/msg00027.html We would also not include all blends there, but select them on an opt-in base. So, if debian-edu is not useful to be installed that we, it shall be removed. At the end, this will reduce the number of selectable blends quite much. I would therefore ask you to rebase your arguments on your experience with the current implementation and not on something that is six months old and not actual anymore. Another point, concerning the argument of "confusing" users: As I said, the blends-tasks package is in place now since eight months, with the current implementation there since ~6 months. Since then there was no single report that someone did not understand the options here -- no bug report, nothing on the installer, blends, or devel mailing lists. I also did an extensive search on the net, and the only thing I found is mentioned in the discussion above and addressed by the changes made after that. Since then, no single problem was reported, with more than 5000 installations according to popcon. This gives a good sign that the addition of the blends to that menu does not confuse people, and I would ask you to show a better empirical evidence that it does. I will not discuss the arguments during the discussion in #758116 here again -- there was a lengthy discussion about this, and the linked postings were covered there as well. It makes no sense to repeat that here again six months later. Concerning your idea of having different install images, I am not convinced that it is a good solution: First, it multiplies the whole image creation process by the number of blends. If we have 10 official architectures and (let's say) 5 blends to be included there, they would then have to manage 60 images instead of 10, with all the requirements that come out of this (installer manual, web page, updates, web space etc.). But it also gives a wrong sign: Debian Pure Blends are by definition integral part of Debian itself. But even now, this is hard to understand for many people -- questions like "what is the difference between Debian Astro and Debian" are quite common, even in front of a poster describing exactly that. With having separate official images for all blends, people would even be more confused. As an example, I would take the Ubuntu approach of having "Ubuntu", "Kubuntu", "Xubuntu" etc. instead of installation options -- people usually think that they have to re-install the system if they want to switch from one flavour to the other. Having similar experience with Debian would be bad for the reputation of the Blends, and for Debian in general. Best regards Ole
Bug#844984: python-astropy: FTBFS: tests failures
Control: forwarded -1 https://github.com/astropy/astropy/issues/5460 I already opened a bug report upstream about this. However, I am not sure whether this may an openssl-1.1 problem with Python; I don't see an obvious bug in the astropy code here.
Bug#840094: blends-dev: Does not recognize multiline dependencies
Hi Andreas, Petter and all, On 15.11.2016 07:09, Andreas Tille wrote: > Hi, > > I just want to announce that I'll be de-facto offline today and > tomorrow. So I can not do further testing of the "Use of uninitialized > value" testing. > > Kind regards > > Andreas. > > On Fri, Nov 11, 2016 at 12:52:46PM +0100, Andreas Tille wrote: >> Use of uninitialized value $_ in pattern match (m//) at >> /usr/share/blends-dev/blend-gen-control line 568, line 28. >> [...] I fixed this in the git. HOWEVER, again: I have no glue what I do here. I just assume that "next unless defined $_" does more or less what it means, and I hope it will do so as well in the line where I put it. Same for "last if !defined $_". I used these two phrases just because they are already somewhere else in the script. Anyone with Perl knowledge: CHECK IT! Cheers Ole
Bug#845145: Please provide CI tests
Package: jsamp Severity: wishlist X-Debbugs-Cc: Paul SladenHi Paul, since jsamp provides some unit tests, it would be useful to run them via the Debian Continious Integration platform, http://ci.debian.net This would ensure that the package keeps in shape when the environment (f.e. the Java version) changes. Best regards Ole
Bug#841610: statsmodels: FTBFS: TypeError: cannot sort an Index object in-place, use sort_values instead
Control: tags -1 patch Upstream there is already a simple patch available for the TypeError: https://github.com/statsmodels/statsmodels/pull/3239 For convenience, it is attached. The IOErrors are gone with the newest Pandas version. However, the ValueError do not disappear when adding tzdata to the build-depends. >From 22cc39f77059297a3ec22d68f1684efd65c433a5 Mon Sep 17 00:00:00 2001 From: thequackdaddyDate: Wed, 16 Nov 2016 14:25:19 -0600 Subject: [PATCH] Simplified commit --- statsmodels/tools/grouputils.py | 1 + 1 file changed, 1 insertion(+) diff --git a/statsmodels/tools/grouputils.py b/statsmodels/tools/grouputils.py index 8e91bba..481ee99 100644 --- a/statsmodels/tools/grouputils.py +++ b/statsmodels/tools/grouputils.py @@ -403,6 +403,7 @@ def get_slices(self, level=0): """ # TODO: refactor this groups = self.index.get_level_values(level).unique() +groups = np.array(groups) groups.sort() if isinstance(self.index, MultiIndex): self.slices = [self.index.get_loc_level(x, level=level)[0]
Bug#844085: Please replace pyfits by astropy.io.fits
Source: pyscanfcs Version: 0.2.3-2 Severity: wishlist Tags: patch Dear maintainer, the package depends on the "python-pyfits" package which is obsolete today and will not see any further development upstream. I am going to create a bug against pyfits asking to remove it from Debian. The "python-astropy" package is the designated successor and provides a drop-in replacement for pyfits. The attached patch does the necessary replacement. Best regards Ole >From 3f4d5181a4ccfe42f50624ab801011991fa001d5 Mon Sep 17 00:00:00 2001 From: Ole Streicher <oleb...@debian.org> Date: Sat, 12 Nov 2016 11:16:12 +0100 Subject: [PATCH] Use Astropy in place of Pyfits --- debian/control | 2 +- .../patches/Use-Astropy-in-place-of-Pyfits.patch | 74 ++ debian/patches/series | 1 + 3 files changed, 76 insertions(+), 1 deletion(-) create mode 100644 debian/patches/Use-Astropy-in-place-of-Pyfits.patch create mode 100644 debian/patches/series diff --git a/debian/control b/debian/control index feb5349..8c53ad6 100644 --- a/debian/control +++ b/debian/control @@ -11,7 +11,7 @@ Build-Depends: python-setuptools, python-matplotlib, python-multipletau, python-numpy, - python-pyfits, + python-astropy, python-scipy, python-wxgtk3.0, imagemagick, diff --git a/debian/patches/Use-Astropy-in-place-of-Pyfits.patch b/debian/patches/Use-Astropy-in-place-of-Pyfits.patch new file mode 100644 index 000..86306b5 --- /dev/null +++ b/debian/patches/Use-Astropy-in-place-of-Pyfits.patch @@ -0,0 +1,74 @@ +From: Ole Streicher <oleb...@debian.org> +Date: Sat, 12 Nov 2016 11:14:23 +0100 +Subject: Use Astropy in place of Pyfits + +Pyfits is obsolete and will not get any upstream updates anymore. It is +replaced by Astropy which provides a drop-in-replacement as astropy.io.fits. +--- + freeze_pyinstaller/debian_ubuntu_bundle_script.sh | 2 +- + pyscanfcs/doc.py | 4 ++-- + pyscanfcs/main.py | 2 +- + setup.py | 2 +- + 4 files changed, 5 insertions(+), 5 deletions(-) + +diff --git a/freeze_pyinstaller/debian_ubuntu_bundle_script.sh b/freeze_pyinstaller/debian_ubuntu_bundle_script.sh +index eaf835c..135b063 100755 +--- a/freeze_pyinstaller/debian_ubuntu_bundle_script.sh b/freeze_pyinstaller/debian_ubuntu_bundle_script.sh +@@ -41,7 +41,7 @@ if ! [ -e $Env ]; then + # Pyinstaller + pip install git+git://github.com/pyinstaller/pyinstaller.git@779d07b236a943a4bf9d2b1a0ae3e0ebcc914798 + # PyFITS +-pip install pyfits ++pip install astropy + fi + source $Env"/bin/activate" + +diff --git a/pyscanfcs/doc.py b/pyscanfcs/doc.py +index 86131b5..ce05cc3 100755 +--- a/pyscanfcs/doc.py b/pyscanfcs/doc.py +@@ -25,7 +25,7 @@ import multipletau + import numpy + import os + import platform +-import pyfits ++import astropy + import scipy + import sys + +@@ -143,7 +143,7 @@ def SoftwareUsed(): +"\n - matplotlib "+matplotlib.__version__+\ +"\n - multipletau "+multipletau.__version__+\ +"\n - NumPy "+numpy.__version__+\ +- "\n - PyFITS "+pyfits.__version__+\ ++ "\n - Astropy "+astropy.__version__+\ +"\n - SciPy "+scipy.__version__+\ +"\n - uilayer "+uilayer.__version__+\ +"\n - wxPython "+wx.__version__ +diff --git a/pyscanfcs/main.py b/pyscanfcs/main.py +index 78c4f80..31e9da1 100755 +--- a/pyscanfcs/main.py b/pyscanfcs/main.py +@@ -57,7 +57,7 @@ import matplotlib.pyplot as plt + import multipletau + + import numpy as np# NumPy +-import pyfits ++import astropy.io.fits as pyfits + from scipy.fftpack import fft + from scipy.fftpack import fftfreq + # SFCSnumeric needs scipy.optimize +diff --git a/setup.py b/setup.py +index 46a8709..baa6337 100644 +--- a/setup.py b/setup.py +@@ -65,7 +65,7 @@ setup( + "matplotlib >= 1.1.0", + "multipletau >= 0.1.4", + "NumPy >= 1.5.1", +-"pyfits", ++"astropy", + "SciPy >= 0.8.0", + "wxPython >= 2.8.10.1" + ], diff --git a/debian/patches/series b/debian/patches/series new file mode 100644 index 000..7b7c503 --- /dev/null +++ b/debian/patches/series @@ -0,0 +1 @@ +Use-Astropy-in-place-of-Pyfits.patch -- 2.10.2
Bug#844087: Please consider removing pyfits
Source: pyfits Severity: wishlist Control: block -1 by 844085 Dear Aurelien, as I wrote on the debian-astro mailing list, I am planning to remove pywcs because it is completely obsoleted by astropy. The same argument is however valid for pyfits, which also will not see any upstream love anymore. IMO the release of Stretch is a good opportunity to force the users into the use of astropy, especially as it is as easy as replacing import pyfits by import astropy.io.fits as pyfits Pyfits has three reverse dependencies (astlib, python-cpl, and pyscanfcs), with the first to being optional (first option there is astropy already). For pyscanfs, I files a bug #844085 asking to switch to astropy. Once they did the warp, would you also consider to remove? IMO that would help to keep Debian clean from obsolete packages. Best regards Ole
Bug#842931: ITP: casacore-data-jplde -- Jet Propulsion Laboratory Development Ephemeris for casacore
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org * Package name: casacore-data-jplde * URL : ftp://ssd.jpl.nasa.gov/pub/eph/planets/ascii/ * License : Public domain Description : Jet Propulsion Laboratory Development Ephemeris for casacore The name Jet Propulsion Laboratory Development Ephemeris are a series of models of the Solar System produced at the Jet Propulsion Laboratory in Pasadena, California, primarily for purposes of spacecraft navigation and astronomy. The package will include the `DE200` and `DE405` tables for casacore. The package is a part of the effort taken by Benda Xu and me within the Debian Astro team to split up the original "casacore-data" package (RFP #761146) by data source into smaller packages which can be maintained independently. The package will be maintained on our git repository under https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-jplde.git Best regards Ole
Bug#842925: ITP: casacore-data-igrf -- International Geomagnetic Reference Field data for casacore
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org * Package name: casacore-data-igrf Version : 10 or 12 * URL : http://www.ngdc.noaa.gov/IAGA/vmod/igrf.html * License : Public domain Description : International Geomagnetic Reference Field data for casacore This package contains the coefficients for the standard mathematical description of the Earth's main magnetic field that is used widely in studies of the Earth's deep interior, its crust and its ionosphere and magnetosphere. The package is a part of the effort taken by Benda Xu and me within the Debian Astro team to split up the original "casacore-data" package (RFP #761146) by data source into smaller packages which can be maintained independently. The package will be maintained on our git repository under https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-igrf.git Best regards Ole
Bug#842935: ITP: casacore-data-tai-utc -- Difference table between TAI and UTC for casacore
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org * Package name: casacore-data-tai-utc * License : BSD-2-Clause Description : Difference table between TAI and UTC for casacore This native package contains the leap second difference between TAI and UTC, created from /usr/share/zoneinfo/leap-seconds.list. The data are in a format specific to casacore. The table is kept in sync with the tzdata package. The package is a part of the effort taken by Benda Xu and me within the Debian Astro team to split up the original "casacore-data" package (RFP #761146) by data source into smaller packages which can be maintained independently. The package will be maintained on our git repository under https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-tai-utc.git Best regards Ole
Bug#842934: ITP: casacore-data-predict -- Earth orientation parameter prediction tables for casacore
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org * Package name: casacore-data-predict * URL : http://maia.usno.navy.mil/ * License : Public domain Description : Earth orientation parameter prediction tables for casacore The IERS Prediction tables provide predictions for the time-varying orientation of the terrestrial reference frame within the quasi-inertial celestial reference frame for casacore. The package will contain the `IERSpredict` and `IERSpredict2000` tables for casacore. The package is a part of the effort taken by Benda Xu and me within the Debian Astro team to split up the original "casacore-data" package (RFP #761146) by data source into smaller packages which can be maintained independently. The package will be maintained on our git repository under https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-predict.git Best regards Ole
Bug#842933: ITP: casacore-data-eop -- Earth Orientation Parameters database for casacore
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org * Package name: casacore-data-eop * URL : https://hpiers.obspm.fr/iers/eop/eopc04/ * License : Public domain Description : Earth Orientation Parameters database for casacore The Earth Orientation Parameters (EOP) describe the irregularities of the Earth rotation with respect to a non-rotating reference frame. The package will contain the `IERSeop97` and `IERSeop2000` tables for casacore. The package is a part of the effort taken by Benda Xu and me within the Debian Astro team to split up the original "casacore-data" package (RFP #761146) by data source into smaller packages which can be maintained independently. The package will be maintained on our git repository under https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-eop.git Best regards Ole
Bug#842936: ITP: casacore-data-observatories -- Table of radio observatory coordinates for casacore
Package: wnpp Severity: wishlist Owner: Ole Streicher <oleb...@debian.org> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org * Package name: casacore-data-observatories Version : 1.0 * URL : https://github.com/casacore/observatory-table * License : Public domain Description : Table of radio observatory coordinates for casacore This package contains a table with radio observatories and their coordinates for the use with casacore. The data is initially taken from Wikipedia, but will be incrementally replaced with verified coordinates. The package is a part of the effort taken by Benda Xu and me within the Debian Astro team to split up the original "casacore-data" package (RFP #761146) by data source into smaller packages which can be maintained independently. The package will be maintained on our git repository under https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-igrf.git Best regards Ole
Bug#842936: ITP: casacore-data-observatories -- Table of radio observatory coordinates for casacore
Hi Jonathan, this package is just one of the standard tables that are available in casacore. Unfortunately, the original table is maintained in a somehow intransparent way at NRAO and ASTRON (looks both institutes maintain it separately), with no "source" available, and without a clear license attached. Also this original list seems to be old (2007?), buggy (f.e. https://github.com/casacore/casacore/issues/304), and incomplete (often some coordinates are missing). To have an open start here, I compiled a list of telescopes and their coordinates from Wikipedia. The list is obviously not very accurate yet, but I hope that we can improve that from public references as time goes. If you can point me to reference geolocations of the VLBI network observatories, that would be really helpful. However, for the effects you describe I am bound to what casacore can take (which is just a plain coordinates in WGS-84 and x/y/z). Best regards Ole On 02.11.2016 13:24, Jonathan Quick wrote: > Hi Ole > > On Wed, November 2, 2016 2:09 pm, Ole Streicher wrote: >> Package: wnpp >> Severity: wishlist >> Owner: Ole Streicher <oleb...@debian.org> >> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org >> >> * Package name: casacore-data-observatories >> Version : 1.0 >> * URL : https://github.com/casacore/observatory-table >> * License : Public domain >> Description : Table of radio observatory coordinates for casacore >> >> This package contains a table with radio observatories and their >> coordinates for the use with casacore. The data is initially taken from >> Wikipedia, but will be incrementally replaced with verified coordinates. > At the highest accuracy, these antenna coordinates are functions of time due > to plate techtonics, with global solutions published at semi-regular intervals > as the ITRF (International Terrestrial Reference Frame). Depending on their > usage within casacore, you may/may not need such accuracies. > > Regards > Dr. Jonathan Quick > VLBI Operations Manager, HartRAO >
Bug#842935: ITP: casacore-data-tai-utc -- Difference table between TAI and UTC for casacore
Hi Jonathan, the original casacore package takes these data from usno; however after some discussion in debian-science I was convinced that tzdata is reliably enough for our use: https://lists.debian.org/debian-mentors/2016/10/msg00494.html Best regards Ole On 02.11.2016 13:37, Jonathan Quick wrote: > Hi Ole > > On Wed, November 2, 2016 2:08 pm, Ole Streicher wrote: >> Package: wnpp >> Severity: wishlist >> Owner: Ole Streicher <oleb...@debian.org> >> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org >> >> * Package name: casacore-data-tai-utc >> * License : BSD-2-Clause >> Description : Difference table between TAI and UTC for casacore >> >> This native package contains the leap second difference between TAI and >> UTC, created from /usr/share/zoneinfo/leap-seconds.list. The data are in >> a format specific to casacore. The table is kept in sync with the tzdata >> package. > Technically the canonical source for this information is the Bulletin C issued > by the IERS at roughly six monthly intervals. I hope the leap-seconds.list is > maintained in reference to those. > > Regards > Jon >
Bug#842934: ITP: casacore-data-predict -- Earth orientation parameter prediction tables for casacore
Hi Jonathan, the package will for sure only contain the prediction for one moment and will be probably outdated when the package reaches the user. However, it will come with a script that can fetch and convert the current data, with the possibility to run this via cron or such. Therefore, the package will have the current data, if the user needs it. Best regards Ole On 02.11.2016 13:34, Jonathan Quick wrote: > Hi Ole > > On Wed, November 2, 2016 2:07 pm, Ole Streicher wrote: >> Package: wnpp >> Severity: wishlist >> Owner: Ole Streicher <oleb...@debian.org> >> X-Debbugs-CC: debian-de...@lists.debian.org, debian-as...@lists.debian.org >> >> * Package name: casacore-data-predict >> * URL : http://maia.usno.navy.mil/ >> * License : Public domain >> Description : Earth orientation parameter prediction tables for >> casacore >> >> The IERS Prediction tables provide predictions for the time-varying >> orientation of the terrestrial reference frame within the quasi-inertial >> celestial reference frame for casacore. >> >> The package will contain the `IERSpredict` and `IERSpredict2000` tables >> for casacore. > These tend to be moving goal posts. Post-analysis of VLBI sessions (and > related space geodetic techniques) gives measurements of EOPs (which are > unknown functions of time due to myriad torques applied to the earth by wind, > ocean, other bodies, human activities etc.) The predictions are just that, an > estimate at a point in time of the future evolution of these parameters which > is subsequently obsoleted by the actual measurements. Hence packaging these > predictions would be of little value, and by their nature, the further into > the future one predicts, the less accurate they become. > > Regards > Jon >> The package is a part of the effort taken by Benda Xu and me within the >> Debian Astro team to split up the original "casacore-data" package (RFP >> #761146) by data source into smaller packages which can be maintained >> independently. >> >> The package will be maintained on our git repository under >> >> https://anonscm.debian.org/cgit/debian-astro/packages/casacore-data-predict.git >> >> Best regards >> >> Ole >> >> >> >
Bug#840473: cfitsio: new upstream version
Aurelien Jarno: > Florian Schlichting: >> a new upstream version of cfitsio was released in April 2016. Would you >> consider packaging it? > > Yes, I am aware of that. Unfortunately it introduces yet another soname > changes, which make things a bit more complicated. Since there is a transition freeze effective from 2016-11-05 [1], this should IMO be addressed ASAP. For the packages under Debian Astro maintenance, I could offer help here. Could you specify where you expect the complications? BTW, I would recommend to put cfitsio under Debian Astro team maintenance, due to the importance of the package. Would you be willing to do that? Best regards Ole [1] https://wiki.debian.org/DebianStretch
Bug#837232: statsmodels: FTBFS: Tests failures
DearYaroslav, could you ensure that the "statmodels" package migrate to testing, so that the dependent packages will not be removed? The package is uninstallable of a number of platforms in the new release, so they should probably removed from testing. Best regards Ole
Bug#840589: Questioning severity of the bug
Hi Peter, could you explain why you think this is of severity "serious"? In my opinion, FTBFS should be "important" as long as there is at least one useful architecture. The definition for important is: "a bug which has a major effect on the usability of a package, without rendering it completely unusable to everyone." which fits to the situation: there *are* still people to which the package is usable (namely the ones which use the architectures where it works). IMO it is up to the maintainer's decision to fix the FTBFS here, or to remove the failing archs from Debian to let the package pass to testing. So, if you not oppose to, I would lower the severity and make it non-RC. Best regards Ole
Bug#840986: Please update to 0.9.0
Package: src:glueviz/0.8.2-1 Severity: wishlist Hi, I am just following the tutorial for glue on the ADASS, and upstream (Tom Robitaille) showed some great new features for 0.9 (mainly 3d). It would be worth upgrading to 0.9, definitely ;-) Cheers Ole
Bug#840589: Questioning severity of the bug
Hi Julien, On 16.10.2016 18:30, Julien Cristau wrote: > Packages must autobuild without failure on all architectures on > which they are supported. Packages must be supported on as many > architectures as is reasonably possible. Packages are assumed to > be supported on all architectures for which they have previously > built successfully. Prior builds for unsupported architectures > must be removed from the archive (contact -release or ftpmaster > if this is the case). > > As skimage no longer builds on architectures where it used to build (and > where it thus is assumed to supported), the "serious" severity for this > bug is correct. Yea, but it is on the decision of upstream and maintainer which platforms are supported by the package. If the maintainer is not able (or willing) to support the platforms that FTBFS, why shouldn't he remove them? Best Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
On 08.12.2016 09:33, Wouter Verhelst wrote: > On Wed, Dec 07, 2016 at 08:59:53AM +0100, Ole Streicher wrote: >> But it also gives a wrong sign: Debian Pure Blends are by definition >> integral part of Debian itself. But even now, this is hard to understand >> for many people -- questions like "what is the difference between Debian >> Astro and Debian" are quite common, even in front of a poster describing >> exactly that. With having separate official images for all blends, >> people would even be more confused. As an example, I would take the >> Ubuntu approach of having "Ubuntu", "Kubuntu", "Xubuntu" etc. instead of >> installation options -- people usually think that they have to >> re-install the system if they want to switch from one flavour to the >> other. Having similar experience with Debian would be bad for the >> reputation of the Blends, and for Debian in general. > > I don't agree with this argument. > > Yes, indeed, in Ubuntu people often don't know that they don't really > need a reinstall to go from Kubuntu to Xubuntu (or whatever), but is > that really a problem? Yes. In the whole 12 years of Ubuntu, they didn't succeed to make clear that [KXG]Ubuntu is not different from Ubuntu except for the package selection. I don't know how important it is for them to keep the unity -- maybe they don't care here much. But for Debian Pure Blends, it is important to have it clear that the Pure Blends are just plain ("Pure") Debian. We don't just use the Debian infrastructure to do something else -- we are an integrated part. DebianMed and Debian Astro are in no way different from Debian, and if you have Debian, then you have the Debian Pure Blends as well: it is just a matter of package selection (and, ofcourse, mainly having the special packages for our fields available). And I still think that it would be horrible, if someone wanting to switch to or from a Pure Blend would feel the need of re-installing. The argument of wanting more than one blend installed at the same time (not so rare within interdisciplinary teams) comes on top of that. > It's certainly *easier* for users to understand that if they want X, Y, > or Z, they just need to install from the X, Y, or Z image. So, if they want KDE, they should just need to install a KDE Debian image? The idea of the Debian Pure Blends is much similar to the idea of the Desktop environments: There is no "KDE Debian", there is "just" a K Desktop Environemnt in Debian. Similarly, the is no "Astro-Debian". There is just a useful environment for astronomers in Debian. And, having extra images per blend would multiply the number of images we have to maintain. Instead of 10 official architectures, we would have 10 + 10 * number_of_blends. Possibly again multiplied by the number of desktop environments, if we apply the same argument to them. I am not sure that the cdimage team would be happy about this. > I don't buy that presenting users with a choice of image to download > "confuses" them, when it in fact *takes away* a very confusing menu from > the installer. I would again ask to present some empirical data that adding the Blends menu is "very confusing". The menu is there since quite a while, and it was presented to many people, and we usually *do* get a response if nobody understands the installation process. So, where are the complaints? After eight months of testing, we can compare the fears here with the collected experience. And this just doesn't support the fears. > I think it's going to be obvious to people that if you > download, say, a Debian Med image, you're going to have a different > experience than if you download a "plain vanilla" Debian image; and > that's *certainly* going to make things easier for Debian Med users, > too. The experience when installing Debian Astro is just Debian. It only adds a useful selection of software on top of that, so that you immediately can start your research, but aside from that everything is as everywhere. So, the difference here is even less than the difference in experiencing different desktop environments. And my experience is that it is easier if people ask about how to install Debian Astro, I can tell "Just select it in the last step of the normal installer", instead of "Go to our home page, download a separate image from there, and install this". It actually makes much difference if the users can "feel" that they are just inside Debian, and not in a special distribution. Best regards Ole
Bug#846197: cpl-plugin-xshoo: FTBFS randomly (tests 126 and 230 fail)
Hi Santiago, OK, I have run it ~15 times without problems last time. I however just disabled the wrong of the two errors (one causes the other), so I switch this now. However, after digging into the code, the failure looks mystic, since the number there is just created from a substring "0.9x11" by taking everything left of the x, and then divided by 3600. The string is fixed in the test. It looks like that there sometimes an additional digit slips in, but in one case (20161207T200220Z) the extraction results in 25000 instead of 0.00025. Really strange. But I don't feel that I can do much here (it is already forwarded). Upstream is informed, however. So, I disabled the (now hopefully right) test, and the package could be built now 40 times in a row without failure. Thumbs crossed :-) Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
On 08.12.2016 18:27, Raphael Hertzog wrote: > - trying to keep blends-tasks now because we have no better option on the > table right now is not a good move either. Had you not circumvented the > d-i review at the time you introduced blends-tasks, then maybe you could > have advertised the limitation of tasksel and we could have found sooner > someone willing to fix this even if nobody in the blends team had the > required skills... The current solution was announced in bug #758116, which at that time was assigned to tasksel, so you will find it in your mail archive. In this bug, there is also a discussion about the limitations of tasksel (starting in 2014, which is probably soon enough!), and also some statements from the d-i team that they will probably not have time to implement something here. There is (and was) no circumvention of a d-i review. The menu is available, and you can always review it and file a bug if it has a problem -- like for any other aspect in Debian. This didn't happen for the blends-tasks package so far. This shows that either the problems were not big enough, or that d-i just had no time to do so. In the latter case, I don't see why d-i would have more time if the tasks menu would be in tasksel-data. In my opinion it *is* a good idea to spread responsibilites; especially when one team is overloaded and the topic fits so well into the field of another team. And I see no advantage to move the responsibility of the blends submenu back from the blends team to d-i. > But more importantly, we need to not show that page at all. I would like > to suggest a first screen: > > Install packages for a: > > [X] standard desktop > [ ] standard server > [ ] minimal server > [ ] Show me more options > > You only see "tasksel" if you check the "Show me more options" which > should be unchecked by default. There's code that translates each option > into default selections at the tasksel level. If this could be implemented in Stretch, it would be a good compromise. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Phil, On 10.12.2016 01:03, Philip Hands wrote: > Just to test things out, if one adds: > > url=hands.com/d-i/bug/846002/preseed.cfg > > to the kernel command line (so, hitting TAB as the installer's boot menu) > it will tweaks d-i to have such a menu. To me, this looks like a very nice solution! In the tasksel screen, the "back" button was enabled for the first time, but produced an error and brought me back to the list of installation steps. Going through the sofware selection a second time made the back button disappear. I have absolutely no experience with preseeding, so I can't test it more than the interactive use case. The advantage of your solution is that one doesn't need to touch tasksel itself, which may address Cyrils doubts. And since Holger also seems to be happy with this solution: maybe we could focus on this in the further discussion? Best regards Ole
Bug#846002: blends-tasks must not be priority:important
Hi Michael, On 10.12.2016 14:48, Michael Biebl wrote: > Am 10.12.2016 um 14:15 schrieb Ole Streicher: >> Hi Michael, >> >> On 10.12.2016 13:18, Michael Biebl wrote: >>> While I have no opninion on whether blend-tasks is important enough to >>> have installed by default, I would urge to revisit the idea to (mis)use >>> the package priority to achieve that. >> >> That is the standard way how tasksel gets its menu: tasksel-data is >> marked as "important" and that way it finds its way onto the installer >> image. So, all arguments you have here are also valid for the default >> tasks. If we think this is wrong, we should get a completely different >> solution here -- which is IMO outside of the scope of this bug. > > Well, two wrongs don't make one right. If we would remove one of them, the other would still stay there, and the problem remains. > An obvious solution seemed to me, to make tasksel(-*) depend on > blends-tasks. This why, the package would be marked as auto-installed, > and should you later decide to implement that differently, say directly > in tasksel, then tasksel can simply drop that dependency. The disadvantage is that people can't uninstall the blends task. And at least Holger clearly indicated that he wants to have this option. > I assume though you have considered this obvious solution and decided > against it for some reason? The major reason is that during the discussion of the blends into tasksel became clear that the tasksel maintainers are already overloaded. Having to manage the blends tasks here would put another work on top of that. Additionally, they have no competence in deciding neither which blends should go in, nor about how to describe them. This all done by the Blends team. So, it seems to be natural for me to have the blends tasks maintained by this team, and not by the tasksel developers. In principle, the same would be applicable for the desktop selection, however there seems to be no dedicated team for the desktop. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Phil, On 10.12.2016 12:06, Philip Hands wrote: > Anyway, having done it, my first impression (which I'm surprised by) is > that the list is too short -- I think that it is perhaps because it is > much easier to select one option from a list than it is to decide what > combination of options one wants. My feeling is exactly the opposite: the having the two most prominent options there is (if they have good names), with the "90% standard" preselected seems simple enough; in the discussion of this bug there was even the proposal (as I remember) to have just *one* option, which also would not be too bad: Most people probably don't care here anyway to select something, and just having *one* checkbox here would give room for a detailed explanation what is going to be installed then. Then, even the "server" would move to the "extended" selection -- servers are usually installed by more experienced users which can deal with more options. In any case, I would like to hear the opinion of Cyril or other d-i people here. Best regards Ole
Bug#846002: blends-tasks must not be priority:important
Hi Michael, On 10.12.2016 13:18, Michael Biebl wrote: > While I have no opninion on whether blend-tasks is important enough to > have installed by default, I would urge to revisit the idea to (mis)use > the package priority to achieve that. That is the standard way how tasksel gets its menu: tasksel-data is marked as "important" and that way it finds its way onto the installer image. So, all arguments you have here are also valid for the default tasks. If we think this is wrong, we should get a completely different solution here -- which is IMO outside of the scope of this bug. > If a package was pulled in by debootstrap because of its priority, it's > marked as manually installed. > This is usually bad, because say later on you decide that you want to > change the name of the package or implement this differently, the > manually installed package sticks around. But this problem is true for all manually installed package, right? So, a manually installed iceweasel would have a renaming problem as well. Couldn't it be solved by a transitional package? I would also consider this as not being release critical. The current solution would also have the advantage that people who actually hate the blends task list (like Holger Levsen: "I will get used to type 'apt remove blends-tasks' as well.") have the possibility to remove that from the tasks selection list. So, this would probably have more acceptance than a solution where the blends tasks are glued into the main task selection. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
On 09.12.2016 08:37, Philip Hands wrote: >> On Wed, 07 Dec 2016, Philip Hands wrote: >>> It could be much improved by making it more obvious that the heading is >>> a heading. Even if we're unable to stop headings having a checkbox, we >>> could change the text and the hierarchy slightly to be something like >>> this: >>> >>> [ ] === Debian Desktop Environments: >>> [x] ... Gnome >> [...] >>> Would that cheer people up without needing a major rewrite of tasksel? This improves the situation, and could probably done quite simple at the same place where the ellipsis (...) is introduced: https://sources.debian.net/src/tasksel/3.38/tasksel.pl/#L360-L366 One just needs to find out here that it is actually a heading. > I think that should be a select, rather than a multi-select: > > --> standard desktop (will install $DESKTOP) <-- > standard server (includes ssh) > other use cases > > (or however the UI presents it) > > The reason for the extra bits in brackets is that I've always thought > the tasksel menu was hiding just a little too much of what was meant by > the options. I would vote for that, however we would need 1. someone willing *and* competent (the latter excludes myself) to implement this in tasksel, 2. someone convincing Cyril that this is ready to go into Stretch: On 08.12.2016 23:20, Cyril Brulebois wrote: > While it's clear to me we need to fix the blends situation at some point > before the release (couldn't find time to do so yet; last resort option > is masking all of them entirely), I'm rather dubious about changing the > package selection/tasksel screen at this point of the release cycle. Volunteers for any of those tasks? Best regards Ole
Bug#846002: blends-tasks must not be priority:important
Hi again, sorry: missed one point: On 10.12.2016 15:07, Ole Streicher wrote: > On 10.12.2016 14:48, Michael Biebl wrote: >> An obvious solution seemed to me, to make tasksel(-*) depend on >> blends-tasks. This why, the package would be marked as auto-installed, >> and should you later decide to implement that differently, say directly >> in tasksel, then tasksel can simply drop that dependency. >> I assume though you have considered this obvious solution and decided >> against it for some reason? If you mean here why we just didn't put blends-tasks as dependency into tasksel-data: this wouldn't help, since the Policy says (§2.5): "Packages must not depend on packages with lower priority values" Since tasksel-data is Priority: important, blends-tasks would need to have this priority as well. Ole
Bug#848112: Python-skimage depends on unavailable package python-dask
package: python-skimage version: 0.12.3-2 severity: serious The Python 2 version of skimage depends on a package "python-dask" that is not available in Debian. There is a patch that make the dependency optional; however the dependency was not removed afterwards. For Python 3, this seems to work. Since skimage is one of the central packages, I would again ask to put it under science|python team maintenance. Especially when under some time pressure (upcoming freeze, combined with autoremovals of packages) it would help a lot if the problems could be debugged within a standard Debian developer workflow, without the need to switch to github or so. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Philip, On 06.12.2016 20:43, Philip Hands wrote: > Could we serve their needs with an extra debian-installer/blend > preseed to deal with this, probably aliased as just 'blend' so that > one could type something like: > > blend=med > > when booting the default media to get the desired result? I think this is really unergonomic, since people need to understand or remember installer boot time options. Boot prompt options are magic for many users, and they need to read the documentation to get this. And it is not recoverable: imagine that someone forgets to put it there or made a typo, he cannot go back and change this -- he has to go through the full installation process again. And it does not really *present* the blends to the user; he already needs to know what is there. > If we then made the ISOs easy to tweak, so that the default option > on the Debian-Med ISOs included blend=med on the command line by > default, would that actually be better than what we have, and also > allow us to drop the problematic tasksel items? Since I already answered this, I hope it is OK to just copy my old argument here: I am not convinced that it is a good solution: First, it multiplies the whole image creation process by the number of blends. If we have 10 official architectures and (let's say) 5 blends to be included there, they would then have to manage 60 images instead of 10, with all the requirements that come out of this (installer manual, web page, updates, web space etc.). But it also gives a wrong sign: Debian Pure Blends are by definition integral part of Debian itself. But even now, this is hard to understand for many people -- questions like "what is the difference between Debian Astro and Debian" are quite common, even in front of a poster describing exactly that. With having separate official images for all blends, people would even be more confused. As an example, I would take the Ubuntu approach of having "Ubuntu", "Kubuntu", "Xubuntu" etc. instead of installation options -- people usually think that they have to re-install the system if they want to switch from one flavour to the other. Having similar experience with Debian would be bad for the reputation of the Blends, and for Debian in general. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Tollef, On 06.12.2016 17:04, Tollef Fog Heen wrote: > ]] Ole Streicher > >> On 06.12.2016 10:37, Holger Levsen wrote: >>> And this *is* still pretty confusing, though admitly better than it was >>> half a year ago. >> >> The current implementation has a popcon of >5000, without a single >> complaint or confusion documented in the web within the last six months. >> This is at least *some* empirical evidence that it is not "pretty >> confusing", and again I would ask you to show any better empirical data >> here to support your own point. > > It's confusing enough that when I've had engineers from a provider > install Debian for me, I have ended up with a desktop rather than server > installation. Should I have filed a bug about it? Maybe. I think you should, since this is the preferred way to let the maintainer know about problems. This is even more true for prereleases like alpha6 ff., since they depend on bugreports to get finished properly. But this is not (completely) what I mean: I didn't only check for bug reports, but also for help requests or comments. And the status is: nobody had so much difficulties with the additional choices, that he asked what to do. So, although the current way has been used by many people, no difficulty was documented (with the exception I already mentioned). > I think it would be better if we moved most of tasksel out of the > installer entirely and had an app store of some sort where applications > and blends could all be better presented with screenshots and > all. I disagree here: the installer is a kind-of assistent which leads you through the installation process. It is much easier, even for me, to follow these steps than to need to remember to start the app store afterwards. BTW, tasksel is able to do both: it is called from the installer, but also max be called afterwards to change the selection if needed. IMO this is the optimal way to interact with the user. It is just the implementation that is hopelessly outdated. > Again, I don't know how feasible it is to end up with a better design > for stretch, but I don't think the current design is particularly > scalable and should be replaced for buster. Fully Ack. I see the current solution to integrate the Blends in Stretch as a compromise for Stretch only; afterwards we should look to rewrite tasksel for a better scalability. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
On 06.12.2016 20:13, Tollef Fog Heen wrote: > Debconf has support for pluggable UI widgets, so somebody could do this > without _too_ much work in the graphical version if they wanted, with > fallback code for the curses and text versions. In principle, you are true. One of the reasons that I pushed the debian-blends-tasks.desc in spring already was that we can see whether this solution is a good compromise, of if we need to do further changes. For the moment, I don't see pluggable UI widgets as an option anymore: this would either mean to rewrite much of tasksel, or to add another step (--> a new program/package) to the installer. With the ongoing freeze, this doesn't sound realistic. For Buster, we should think about a significant tasksel change, however. Best regards Ole
Bug#846197: cpl-plugin-xshoo: FTBFS randomly (tests 126 and 230 fail)
Control: forwarded -1 PIPE-6911 Control: tags -1 unreproducible Hi, I forwarded the problem to ESO, for reference it gets there the issue number PIPE-6911. Their ticket system is however internal only, so that I can't give an URL here. I also tried to reproduce the problem, but was not successful: More than 10 builds (cowbuilder, x86_64, stretch) in a rowwere OK, without a single failure.The same for sid. Therefore, I also do not expect that upstream can do much here. I will disable the test in the moment, but re-enable it when there is a new upstream version. Please re-open if you then still encounter the problem. Best regards Ole
Bug#847970: sunpy FTBFS in stretch due to missing skimage
Hi Adrian, I do not completely understand what the goal of this bug is: The missing build dependency is already noted in the package, and an autoremoval is filed (and sent out to me as the maintainer). So, I am already noted by the problem, and I don't see what the additional bug gives here except the need to close one RC bug more. Wouldn't it be simpler just to mark #840589 as "affects: src:sunpy"? Best regards Ole On 12.12.2016 17:59, Adrian Bunk wrote: > Source: sunpy > Version: 0.7.4-1 > Severity: serious > Tags: stretch sid > Control: block -1 by 840589 > > node-yargs build-depends on python{,3}-skimage, > which is not in stretch due to #840589 >
Bug#850966: Please support ftp search with version number only in the directory name
Package: devscripts Severity: wishlist User: devscri...@packages.debian.org Usertags: uscan I have the following sample download URL ftp://andromeda.star.bris.ac.uk/pub/star/stilts/v3.0-9/stilts_src.zip Corresponding Debian version number should be 3.0.9. There is currently no way to get this downloaded via uscan. I tried version=3 options="uversionmangle=s/\-/./,filenamemangle=s/\/$/.zip/" \ ftp://andromeda.star.bris.ac.uk/pub/star/stilts/v([\d\.\-]+)/ stilts_src.zip but it doesn't work; basically it adjusts upstream version to be 1: uscan info: Newest upstream tarball version selected for download (uversionmangled): 1
Bug#850966: Please support ftp search with version number only in the directory name
Hi Ben, Am 12.01.2017 um 09:22 schrieb Ben Finney: > Ole Streicher <oleb...@debian.org> writes: > How would you expect this to work? one way could be: uscan could check the dir ftp://andromeda.star.bris.ac.uk/pub/star/stilts/ for all dirs matching v([\d\.\-]+), compare with d/changelog (by using uversionmangle), search for the most recent version between the newer ones, then download from that dir the file specified in the second part (stilts_src.zip), and rename it using fileversionmangle (so to v3.0-9.zip in the example). > You're right. The only information UScan can use, is what information is > in the response document. It needs a *single* document with *all* the > relevant versions, as links in the document. if the second component (the file name) does not contain a grouped regexp, it could just be added to the directory link in the first document. Another way would be to use only one URL: ftp://andromeda.star.bris.ac.uk/pub/star/stilts/v([\d\.\-]+)/stilts_src.zip If there is a directory component *after* the grouped regexp, split it into three parts: * Base URL for directory, ftp://.../star/stilts/ * Regexp for the directory search and version extraction, v([\d\.\-]+)/ * suffix path to add for download, stilts_src.zip > How would you propose UScan should detect all the relevant versions? > What URL will respond with a document containing the complete list of > versioned links? See above. Does this make sense? Best regards Ole
Bug#846002: Debian Installer Stretch RC 1 release
On 15.01.2017 05:21, Cyril Brulebois wrote: > The Debian Installer team[1] is pleased to announce the first release > candidate of the installer for Debian 9 "Stretch". > > > Important changes in this release of the installer > == > > * [...] > * As noted in the Stretch Alpha 6 release announcement, Debian Pure >Blends appeared in the Software selection screen. Unfortunately, >concerns voiced back then weren't worked on until after the freeze >started, and a freeze isn't the time where critical screens should >be revamped. Support was disabled accordingly. Since this is still an open discussion in #846002, I would have preferred if you would not try to force your own preference here before the CTTE made its decision. IMO your solution is not in any way cooperative and tries to make the CTTE decision useless. I therefore would ask the CTTE to make a final decision about the inclusion of the blends selection in the Stretch installer. In principle these are *two* decisions: 1. Appearance of the blends in the installer: Please decide, whether * the blends shall be shown in their current (alpha-8) form * The installer is extended to show the desktop and the blends only optionally (as propagated by Phil, and opposed by Cyril) * the blends should not appear in the Stretch installer at all. 2. Maintenance of the blends tasks appearing in the installer: * in a separate package maintained by the blends team * integrated into tasksel and maintained by d-i * be removed completely from the installation process Best regards Ole
Bug#851437: astroplan: FTBFS (requires Internet to build)
The problem here is that with astropy-1.3, doctests are buggy: https://github.com/astropy/astropy/issues/5670 A workaround currently is to disable doctests during the build: python$* setup.py test --skip-docs -vv --args -v Cheers Ole
Bug#846363: Please let casacore-data-tai-utc migrate
Package: release.debian.org Severity: normal User: release.debian@packages.debian.org Usertags: britney The package casacore-data-tai-utc is currently stuck in unstable because of an unsatisfiable Depends: excuses: * 12 days old (needed 10 days) * casacore-data-tai-utc/i386 unsatisfiable Depends: python3-casacore * Valid candidate The package is useful on (some) other archs, however, and will become useful on i386 as soon as the FTBFSs are fixed. Therefore I would ask to force the package migration. Thanks Ole
Bug#846539: ITP: gwcs -- Tools for managing the World Coordinate System of astronomical data
Dear Miguel, would you mind to put the packaging under Debian Astro team maintenance? This would make contributions of others easier, and we would also help you with questions and finally sponsor the package when it is ready. On 01.12.2016 23:56, Miguel de Val-Borro wrote: > * Package name: gwcs > Version : 0.5.1 > Upstream Author : Nadia Dencheva> * URL : https://github.com/spacetelescope/gwcs > * License : BSD-3-Clause > Programming Lang: Python > Description : Tools for managing the World Coordinate System of > astronomical data > > GWCS supports a data model which includes the entire transformation pipeline > from input coordinates to world coordinates. The goal of the package is to > provide a flexible toolkit which is easily extendible by adding new transforms > and frames. > > The package will be maintained using a git repository on alioth. The Debian Astro web page is https://blends.debian.org/astro I am forwarding this to our mailing list as well: https://lists.debian.org/debian-astro Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi all, On 06.12.2016 00:29, Don Armstrong wrote: > [The screen shot Holger linked to is a screen shot of the installer at > the tasksel screen showing an entry for "Debian Blends" followed by a > series of entries which start with leading periods followed by entries > like "HamRadio" and "DebiChem".] This screenshot is outdated since several months. Currently, the wording is different, and the number of included blends has shrunken a lot. Best regards Ole
Bug#846002: blends-tasks must not be priority:important
On 06.12.2016 10:18, Philip Hands wrote: > it also buries the 'standard system utilities' item in the middle of > the list, where it makes even less sense than it did at the end. The positions of the items can be changed by assigning a "Relevance" (one digit) to them. So, if the "Standard" task should be on top (what I would prefer), it should get a high relevance; if it should be at the end, it should get a low relevance. Currently, there is no relevance specified for the "Standard" task, therefore it defaults to 5. The blends tasks have "Relevance: 8", since I intended to have the blends at the end (which is the right place for special tasks, IMO). There is room for one relevance below, and if this is not enough, we could rise the blends relevance: I just thought that there is probably more need to sort things before than after the "Special tasks". Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
On 06.12.2016 10:37, Holger Levsen wrote: > And this *is* still pretty confusing, though admitly better than it was > half a year ago. The current implementation has a popcon of >5000, without a single complaint or confusion documented in the web within the last six months. This is at least *some* empirical evidence that it is not "pretty confusing", and again I would ask you to show any better empirical data here to support your own point. > and it will only get worse, if we would keep it this way… We have *many* more > blends in Debian… Debian Parl, Debian Junor, Debian Edu come to my mind > immediatly. > why list some and not some others? The "single checkbox" is not usable for all blends, since it requires a "one size fits all" solution. For Debian Edu, you already stated yourself that it is useless. Debian Junior and Debian Parl didn't opt in. I therefore don't see a danger that the list will grow endlessly before Stretch. > and that this bug should not be about this tasksel menu but *about this > was implemented*, which is by forcing an unneeded package on each and > every Debian system under the sun. (priority: important…) I already answered to this: there is no technical reason to have the blends task in an individual package; technically it could also be put into tasksel-data itself. However, this would require action from the tasksel team every time something changes in the blends task. d-i is already overloaded, and it will not help if we put another responsibility on top of that. So, it is a good solution to separate the responsibility of the "Special tasks" item to the blends team. Compared to an integrated (into tasksel-data) solution, the size impact is minimal: mainly the overhead of having an additional package there. If we care about this overhead, the problem would apply to tasksel-data as well. Best regards Ole
Bug#846002: Lowering severity
Control: Severity -1 normal Since no objections against my proposal were expressed for a week, I am lowering the severity. Since there is no update of the bug report with more recent experiences, I will to close it as of version 0.6.94 within a few days. Best regards Ole
Bug#846002: Lowering severity
Hi Holger, On 05.12.2016 13:46, Holger Levsen wrote: > I'm sorry that I failed to respond yet. I am quite angry about this: You basically opened this bug by stating that you will do an NMU within 4-5 days, but you already knew that you would not have time to discuss the bug before you planned this to happen. This puts quite some pressure to us to respond within a reasonable time -- and then you (who clearly expressed how important you think this is) hide yourself and fail to respond. This all happens after half a year of silence from you, where we could have discussed this and find alternatives without too much pressure. And you even didn't check the current status before filing the bug. Sorry, but this doesn't sound like a sensible behavior to get a good solution. > Still, this doesnt make a policy violation a non-policy violation. > (Priority: important is *wrong* for this package.) It is as wrong as for tasksel-data: if we want to have blends selection in the installer, then this information needs to be available there. Lowering the bug severity was done since I proposed so and did not get a veto reply (covering my arguments) from you within a week. > You are forcing the installation of blends-tasks on every Debian > system. This is *not ok*. In principle, the whole package could be integrated into tasksel-data. However, this makes the work more complicated for the (already overloaded) tasksel maintainers, since they also would have to deal with the integration of the individual blends. The current solution keeps the maintenance of the blends tasks in the installer in the responsibility of the blends team. I don't see why this solution is worse than to have it combined in a single package (or file) -- it is even better since it moves responsibilities away from an already overloaded team. > I will try to reply ASAP, but… I might fail and just reassign to > tech-ctte. (I also find it very hard to find time for this as you failed > to understand the problem with the current implementation from the > beginning, so I'm doubtful whether trying again (for like the 4th or 5th > time will change anything.) I brought some arguments. It would be nice if you could respond to them. Since you even didn't check the current state, it is hard to discuss here. As I explained, your arguments are already handled with version 0.6.95. I would prefer to have the arguments discussed first instead of just re-assigning to tech-ctte (which you did while I writing this reply, and before you even started to handle Andreas' and my arguments). Best regards Ole
Bug#761146: Re-ITPing
Control: retitle -1 ITP: casacore-data -- Data for Common Astronomy Software Applications core library Control: owner -1 Ole Streicher <oleb...@debian.org> To keep the casacore data packages together, I plan to use the original package as a metapackage that depends on the individual ones, currently: * casacore-data-tai-utc * casacore-data-irgf * casacore-data-lines * casacore-data-observatories * casacore-data-sources Once the remaining packages are there, they will be added as dependencies as well: * casacore-data-eop (ITP #842933) * casacore-data-jplde (ITP #842931) * casacore-data-predict (ITP #842934) but this will not happen for stretch: They are not buildable with the current version of casacore due to https://github.com/casacore/casacore/issues/537 Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
On 06.12.2016 12:02, Philip Hands wrote: > Ole Streicher <oleb...@debian.org> writes: > >> On 06.12.2016 10:37, Holger Levsen wrote: >>> And this *is* still pretty confusing, though admitly better than it was >>> half a year ago. >> >> The current implementation has a popcon of >5000, without a single >> complaint or confusion documented in the web within the last six months. >> This is at least *some* empirical evidence that it is not "pretty >> confusing", and again I would ask you to show any better empirical data >> here to support your own point. > > You seem to be mixing up two things here. > > Holger was saying that the menu is confusing (as am I). > > You're saying that because 5000 people seem to have ended up with this > package on their systems, they were not confused. I am saying that from these 5000 people nobody asked for help or complained in the web, except for the initial wording. This initial documented confusion shows that we actually *get* a response here (and the search is not just useless). > Looking at the graph for debian-astro, is seems to follow a similar > curve, so perhaps these are the people that are installing the > blends-tasks (BTW is there an easy way to check which packages are > installed together via popcon?) The relevant package from debian-astro (which is going to be installed with by tasksel) is "astro-all". It has a popcon of 148, so most of the people installing blends-tasks (5400) then do not install debian-astro. The curve for both is similar, since both were introduced together: astro-all was specifically created to do the actual installation of Debian-Astro via the installer tasksel. > There is no need for them to tick the 'Special tasks' menu item in order > to install any of the the blends, so to some tiny extent they were > confused when they did that, were they not? I also find the structure here not optimal; this is however given by the limitation that tasksel uses debconf, which has only limited possibilities. The checkbox in "Special tasks" is confusing for sure. However, this does not add confusion: the same problem is already there (and was so in Jessie) with the "Desktop environment", so people need to get used to it anyway -- we don't solve this by deciding the current issue. What the empirical search however shows that the blends-tasks didn't add additional confusion here. Best regards Ole
Bug#849502: Use of embedded external libraries
I removed the usage of six, configobj, ply, jquery and jquery.dataTables in the git repository. However, I still cannot remove the use of the local pytest, since astropy doesn't test successfully with the current version in Debian (3.0.5). See also https://github.com/astropy/astropy/issues/5277 https://github.com/astropy/astropy/issues/5670 Once this is fixed, I will remove that as well. It will be switched off in astropy/tests/helper.py then.
Bug#849196: Sometimes, supress_warnings misses one of its attributes
Hi Sandro, On 30.12.2016 15:01, Sandro Tosi wrote: > On Fri, Dec 23, 2016 at 9:47 AM, Ole Streicher <ole.streic...@gmx.de> wrote: >> This is a regression; it did not happen with 1.11. Please fix this >> regression ASAP so that skimage can migrate safely before the freeze. > > as asked on the github issue, is disabling parallel tests execution in > skimage a viable temporary solution? For the moment, I disabled the build time tests in skimage completely because of #849196, to ensure it migrates before the freeze (I took the really latest chance to do so). Therefore, I would not touch the package before Jan 5. Once it is in testing, It would be however very important to re-enable the tests, since this is more a crude hack than a clean solution. Unfortunately, I am not familar with the skimage code (I just did the firefighter job here); specifically I don't know where to switch off parallel tests, and what other implications that has. Maybe, you discuss this with Yaroslav (in Cc), who is the skimage package maintainer? Also the parallel bug I opened on the "scikit-image" github (referenced on the numpy issue) may help there. Finally, Yaroslav should decide here. I personally however would much more prefer to go back to the latest numpy 1.11 in testing, and update only after 1.12 is released. I very dislike having workarounds because a beta or a release candidate uploaded to testing is buggy. And I would also prefer not to have a beta or RC shipped with Stretch. Best regards Ole
Bug#848859: FTBFS randomly (failing tests)
Hi all, On 04.01.2017 20:57, Santiago Vila wrote: > I still want to build all packages and have 0 or 1 failures, > so in this case the probability should be 1/50/2, i.e. 1%. > > I think this is still feasible. My experience is that the buildds that are currently in use provide more build problems than the packages themself. BTW, why don't you count this as RC? [statistical] >> The "fix" for such cases is the increasing of the threshold or disabling >> the test completely. Because you can do nothing with it due to the >> nature of numerical simulations. Just disabling would be bad, since the test still can point to some problem in the implementation, like optimization problems or such. IMO the correct solution here would be to remove the randomness and to seed with a fixed value (which is known to give a result within the expectations). > But as far as there are people in this project who consider that a > package which FTBFS on single-CPU machines more than 50% of the time > is ok "because it does not usually fail in buildd.debian.org", > we are doomed. We have 10 archs, and with a 50% chance of failure you will not get any version built. Even when the buildds try it two or three times. > The problem I see with this threshold thing is that every maintainer > seems to have his own threshold, different from the others. > > In case we decide about RC-ness depending on probability of failure: > What threshold do you think we should use for a single package and why? > > [ I say that 1% of failure is the maximum we should allow, and I've > explained why, but I would love to hear your opinion on this ]. I would not put any direct number here, but be pragmatic: We need to build the package from source, and this has to be done on all supported architectures, and on the buildds. As long as this is done, I see no reason for an RC bug. If a package fails on a supported arch on the buildds, this is RC, independent of whether this came from an architectural difference or from a random build failure. Your tests with repeated builds help the maintainer to find the cause in that case, and for this they are helpful. But not release critical by themself. Cheers Ole
Bug#848859: FTBFS randomly (failing tests)
Hi Santiago, On 04.01.2017 01:41, Santiago Vila wrote: > On Tue, 27 Dec 2016, Ole Streicher wrote: > Hello Ole. Thanks for your reply. Please don't forget to Cc: me if you > expect your message to be read. OK, however I usually assume that a bug submitter actually reads the messages for his submitted bugs. >>> In particular, if something happens 1 every 20 times on average, the >>> fact that it did not happen when you try 10 times does not mean in any >>> way that it may not happen. >> >> >> I must however say that I don't see why a package that fails to build >> once in 20 builds would have a release critical problem: > > It's in Release Policy: Packages *must* autobuild *without* failure. > > If a package fails to build from time to time, that's a failure. Packages actually *do* fail from time to time, when I look into my autobuilder. Not due to the package, but due to glitches within the buildd infrastructure. Would you consider this a failure? >> I totally agree that catching random failures >> is a good quality measure, but this is IMO severity "important" at maximum. > > Well, would you say it's RC if it fails 99% of the time? > I guess you would. I would consider a bug RC if it actually doesn't build on our buildds. >> And it would be nice to have this QA test not just before the release, >> but already early in the release cycle. This would help a lot to avoid >> stressful bugfixes just before the freeze. > > Well, if you see my list of FTBFS-randomly bug here: > > https://bugs.debian.org/cgi-bin/pkgreport.cgi?tag=ftbfs-randomly;users=sanv...@debian.org > > you will see that I started to report FTBFS-randomly bugs several > months ago, not last week, and not last month. > > What would really help is to have somebody else reporting these bugs, > because apparently very few people want to report them. > > Also, you can't seriously "complain" that a bug reporter didn't not > report something earlier! Of course that it would be nice to do this > kind of QA stuff at every point in the release cycle, but as the same > software that we maintain, my bug reports come with no warranty, not > even the warranty that the bug is reported "as soon as it exists"! :-) What I do complain about is that doing this just before the release adds additional pressure to the package maintainers: The problem for us ordinary people is that most of the time my packages look fine, without serious problems, and just before the release all those RC bugs pop up in bunches, with quite some time pressure to solve them. If the RC bugs were randomly distributed over time, it would be more time to solve them, and the quality of the fixes would be better: for example, in the moment, I would just disable build time tests that randomly fail, while usually I would try to see why they fail and fix that. In the moment, there is no time for this. Doing release QA just before the release leads to quick hacks to keep things there, while a continious QA really solves them. Best regards Ole
Bug#848859: FTBFS randomly (failing tests)
On 05.01.2017 11:36, Santiago Vila wrote: > On Thu, Jan 05, 2017 at 11:06:50AM +0100, Ole Streicher wrote: >> Hi all, >> >> On 04.01.2017 20:57, Santiago Vila wrote: >>> I still want to build all packages and have 0 or 1 failures, >>> so in this case the probability should be 1/50/2, i.e. 1%. >>> >>> I think this is still feasible. >> >> My experience is that the buildds that are currently in use provide more >> build problems than the packages themself. BTW, why don't you count this >> as RC? > > Can you clarify the question? I don't understand what you refer exactly. It does not make sense to have 100% reliability on the package itself to build but only 95% on the buildds. If we want to have reliability, then every chain member counts -- and the weakest one in the moment for me does not seem to be the packages themselves, but the buildd setups we currently have. So, if you consider unreliable package builds as an RC problem, you should also consider the unreliable buildd setup as an RC problem. >>> But as far as there are people in this project who consider that a >>> package which FTBFS on single-CPU machines more than 50% of the time >>> is ok "because it does not usually fail in buildd.debian.org", >>> we are doomed. >> >> We have 10 archs, and with a 50% chance of failure you will not get any >> version built. Even when the buildds try it two or three times. > > It's actually more subtle than that. > > The package may be ready to be built on multi-core machines, but not on > single-core machines. So, the probability that I measure (greater than 50%), > might not be the same as the probability on official autobuilders. Ususally it is the other way around: packages build fine single-threaded, but don't do so on many threads (or CPUs). At the end it it also depends on how many CPUs you actually get for the build -- if there are some packages built in parallel, the result may be as for a single CPU. > Then there is the other subtle thing: The package may be Arch:all, in > which case a 50% of probability (this time independent on the number > of CPUs) may go unnoticed very easily, either because the maintainer > uploaded the .deb him/herself, or, if the maintainer uploaded > in source-only form, we can just be lucky for that time. I would favour much to ignore any prebuilt binaries and to re-build everything anyway. > We provide software which is free to be modified. If people is going > to modify it and then it fails because the package only builds ok in > buildd.debian.org, then we are removing one of the essential freedoms, > the freedom to modify it, and the package becomes de-facto non-free, > even if we still distribute it in main. First, the multi-CPU buildd setup is free (as in speech), isn't it? So, depending on this does not make anything non-free. I don't see why the dependency on a specific build environment would change the DFSG freeness at all (as long as the build environment itself is DFSG free). Then, we are speaking about random failures. If you use a single-core and see a build failure, you can just pragmatically repeat it. Sorry, I don't see why this would restrict DFSG freedom here. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Cyril, On 20.12.2016 10:59, Cyril Brulebois wrote: > dropping blends entirely is still my default option in case proposed > changes have too far reaching consequences. As I already wrote several times: before you do so, please show some evidence that in the half year that the current version of the installer containing a blends selection has added an unacceptable amount of confusion and that we can't solve that by changing the wording or such. We already have more that 5700 popcon-counted installations with the blends selection in the installer. This should give some base for that. And at least by searching the net, I could not find anything. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Sam, On 21.12.2016 16:10, Sam Hartman wrote: >>>>>> "Ole" == Ole Streicher <oleb...@debian.org> writes: > I don't find quoting popcon stats useful. You've used them to support > the claim both that this is important and that users don't find it > confusing. I am quoting popcon here since they give a lower estimate of the number of users who actually did the test. Nothing more. Nothing about importance. > I think your reasoning rests of the false assumptions that users are > likely to report bugs when they find something confusing. > In my experience that's not the case. Instead, they are likely to get > grumpy and decrease their overall confidence in some software. > Users cannot be counted on to proactively report confusing aspects of > software. I don't speak about bug reports only. I have searched the net for any appearance of problems with the blends; also discussion forums and such. And as I wrote in one of my early statements: there *was* such a case: in a german forum, someone asked "what does this blends stuff mean?". This shows that one *gets* a response here. After that response, we changed the wording; no further complaints were found after that. And I would also not just count "end user reports". If for example someone would have done an install party (or such), he would know what the usual questions of users were. Also if someone recommended to install Stretch to his friend and had to help somewhere. He could (and should!) then do a bug report. Didn't happen yet. "Confusion" is not just something mythical that we can't see empirical. We will see it in help requests, blog posts, also bug reports, and other complaints. If the raise of confusion is "inacceptable" as stated here, I would really ask for some empirical evidence for this. At least, I did some homework by checking that no complaints popped up somewhere in the net (except the one I already mentioned). Best regards Ole
Bug#848758: Latest upgrade of Numpy breaks tests of other packages (Was: Bug#848758: python-skbio: FTBFS: Test failures)
On 21.12.2016 16:17, Adrian Bunk wrote: > I can clearly understand why Andreas is unhappy to see when such a > change is uploaded without prior warning 1.5 months after the start > of the freeze, and 1 week before an important deadline for his package. Sure; what I also don't understand is why numpy pushes its RC and betas into unstable instead of experimental (and then maybe check or asks for checking for the reverse deps). This makes it harder to revert if there are problems. Sandro, maybe you could explain that a bit? Cheers Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Holger On 20.12.2016 15:27, Holger Levsen wrote: > On Tue, Dec 20, 2016 at 03:16:50PM +0100, Ole Streicher wrote: >> Again: the installer is there to test for 6 months now, but if it is >> inacceptably bad: why are there no complaints? > > the complaints have been there for months, you just choose to consider > them invalid. some people dont like to repeat themselves. After six months of testing, I would expect that the fears that were expressed at that times were supported by some real solid experiences. This is a big difference and in no way a repetition. Ole
Bug#848758: Re: Bug#848758: Latest upgrade of Numpy breaks tests of other packages (Was: Bug#848758: python-skbio: FTBFS: Test failures)
Hi all, On 21.12.2016 11:59, Andreas Tille wrote: > On Wed, Dec 21, 2016 at 11:40:16AM +0200, Adrian Bunk wrote: >> For python-skbio it is *really* time to panic *right now*. > Thanks for confirming that I was not actually panicing. ;-) While I agree in principle, I would like to remind the following as well: The current FTBFS mainly (as far as I can see) come from a change that was announced already two numpy versions ago: that numpy will refuse to accept floats as indices. The according functions are since then marked as "deprecated" and issue a warning. Numpy introduced this change with 1.11rc already, which was uploaded to Debian about 6 months ago. After it appears that many packages (als in Debian) have problems with this, they postponed it for another release -- which happens just now. I am a bin angry here about unresponsive upstreams, which just ignore these changes as long as possible, and I am not sure whether we should ship Stretch with an outdated numpy release just because there are a number of upstreams that are too lazy to keep their packages up to date. I would really ask to push upstream hardly to do the required changes; maybe we can decide if it is needed to revert that later. The changes are required anyway; doing them sooner than later is no mistake in any case. (I am too lazy in the moment to support my statement with links; if you need, I will do so ofcourse) Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Steve, On 20.12.2016 15:16, Steve McIntyre wrote: > I *have* also seen users confused by the addition of the blends Can you be more specific here? Old wording (with many blends) or current solution? What was the specific problem? > into the tasksel list. A better split of the tasks (like Phil is > suggesting) would help a lot here. This is for sure. Cyril just states that he rather would love to remove the blends completely, and this is something I am arguing against. That Phils solution is a great compromise, is out of question. IMO it would help much if d-i would help here a bit instead of just trying to play a veto game. Best regards Ole
Bug#846002: blends-tasks must not be priority:important (was Re: Bug#846002: Lowering severity)
Hi Cyril, On 20.12.2016 15:01, Cyril Brulebois wrote: > Ole Streicher <oleb...@debian.org> (2016-12-20): >> As I already wrote several times: before you do so, please show some >> evidence that in the half year that the current version of the installer >> containing a blends selection has added an unacceptable amount of >> confusion and that we can't solve that by changing the wording or such. > > Having more choices means more confusion. Look at any UX studies, or > install parties. > > Also, choices aren't consistent depending on installation media, which > doesn't help. > > You might call it me being weird, but this is how we've tried to balance > choices in D-I for a few years (10+ AFAICT), as already confirmed by > Christian earlier. The addition of various blends in the path of all > users shifts this balance, in the wrong direction. And by how much? If it is just a very little bit, this could be acceptable. It seems clear by the whole discussion that tasksel is *already* quite confusing to the users in an unacceptable way, and that we should change it in buster. So, whatever we do now is just the last step in a dead end. BTW, adding LXQT to the desktops also goes into the wrong direction, since it adds another choice. This shows, that you don't take the "wrong direction" argument serious yourself (resp. Christian, who did the patch in the tasksel git). >> We already have more that 5700 popcon-counted installations with the >> blends selection in the installer. This should give some base for >> that. > > Surely, people asking for blends are using blends selection. That's nice > and I'm pretty sure nobody is going to dispute that. But that shouldn't > make d-i more confusing for others. So, please *show* that the current solution does add confusion in an unacceptable way. Show for example installation reports. Or something else which shows clearly that the current concrete solution in not acceptable. Based on the concrete experience of the current installer. And, I didn't just search through the astronomy related pages for issues, but tried to catch any report for the new installer -- and didn't find anything. Again: the installer is there to test for 6 months now, but if it is inacceptably bad: why are there no complaints? Best regards Ole
Bug#848758: Latest upgrade of Numpy breaks tests of other packages (Was: Bug#848758: python-skbio: FTBFS: Test failures)
On 21.12.2016 19:16, Sandro Tosi wrote: >> They are still, what the name suggests: candidates, with no confirmation >> to be useful in a production environment. I don't see why they should >> ever migrate to testing (as they did in 1.11.1rc.1). Last time, we had >> an numpy RC in testing for more than four months (2016-05-06 to >> 2016-10-16)! This doesn't confirm a sigh quality tendency. And imagine > > this is the diff between 1.11.2rc1 and 1.11.2 (not the same versions > you mentioned, but the most recent one we can make such comparison): > > https://anonscm.debian.org/cgit/python-modules/packages/python-numpy.git/diff/?h=upstream/11.11.2 ... and the difference to the one before is larger, and more critical, and longer in testing. > the version is just a name, there will be bugs even in the most shiny > new release of every software An "RC" instead of a release is there for a reason. The last (1.11) numpy release process has shown that. It would be nice if the Debian packaging would reflect this difference. Best regards Ole
Bug#848758: Latest upgrade of Numpy breaks tests of other packages (Was: Bug#848758: python-skbio: FTBFS: Test failures)
Hi Sandro, On 21.12.2016 16:43, Sandro Tosi wrote: > On Wed, Dec 21, 2016 at 10:27 AM, Ole Streicher <oleb...@debian.org> wrote: >> Sure; what I also don't understand is why numpy pushes its RC and betas >> into unstable instead of experimental (and then maybe check or asks for >> checking for the reverse deps). This makes it harder to revert if there >> are problems. >> >> Sandro, maybe you could explain that a bit? > > early/wider exposure, and they tend to be of high quality. They are still, what the name suggests: candidates, with no confirmation to be useful in a production environment. I don't see why they should ever migrate to testing (as they did in 1.11.1rc.1). Last time, we had an numpy RC in testing for more than four months (2016-05-06 to 2016-10-16)! This doesn't confirm a sigh quality tendency. And imagine that happens now; chances are high that we release Stretch with a numpy release candidate. IMO the way numpy changes are to be tested in Debian needs some adjustment. Best regards Ole
Bug#848112: Python-skimage depends on unavailable package python-dask
Control: tags 848112 pending Hi all, On 23.12.2016 10:08, Andreas Tille wrote: > Ole, if you would please commit the patch you named in #848112 and > upload the package to make sure it can migrate right in time and will > enable other packages to migrate as well? > > HOWEVER, we have another problem: If I build the current state in Git > I'm running into: > TypeError: 'numpy.float64' object cannot be interpreted as an index I have a fix for both, and after a final test build, I will upload. The package still seems to be in bad shape; during the docs generation exceptions are thrown like ValueError: 3-dimensional arrays must be of dtype unsigned byte, unsigned short, float32 or float64 TypeError: 'numpy.float64' object cannot be interpreted as an index They are ignored, however, but show that some work is still needed. Cheers Ole
Bug#849196: Sometimes, supress_warnings misses one of its attributes
Package: src:python-numpy Severity: serious Version: 1:1.12.0~b1-1 X-Debbugs-Cc: debian-scie...@lists.debian.org Forwarded: https://github.com/numpy/numpy/issues/8413 Affects: src:skimage When trying to compile skimage, I sometimes get the following error: Traceback (most recent call last): File "/usr/lib/python2.7/dist-packages/nose/case.py", line 197, in runTest self.test(*self.arg) File "/build/skimage-0.12.3/debian/tmp/usr/lib/python2.7/dist-packages/skimage/transform/tests/test_integral.py", line 46, in test_vectorized_integrate assert_equal(expected, integrate(s, r0, c0, r1, c1)) # test deprecated File "/build/skimage-0.12.3/debian/tmp/usr/lib/python2.7/dist-packages/skimage/transform/integral.py", line 86, in integrate warn("The syntax 'integrate(ii, r0, c0, r1, c1)' is " File "/build/skimage-0.12.3/debian/tmp/usr/lib/python2.7/dist-packages/skimage/_shared/_warnings.py", line 16, in warn warnings.warn(message, stacklevel=stacklevel) File "/usr/lib/python2.7/dist-packages/numpy/testing/utils.py", line 2199, in _showwarning self._orig_show(message, category, filename, lineno, AttributeError: 'suppress_warnings' object has no attribute '_orig_show' or ERROR: test suite for -- Traceback (most recent call last): File "/usr/lib/python3/dist-packages/nose/suite.py", line 229, in run self.tearDown() File "/usr/lib/python3/dist-packages/nose/suite.py", line 352, in tearDown self.teardownContext(ancestor) File "/usr/lib/python3/dist-packages/nose/suite.py", line 368, in teardownContext try_run(context, names) File "/usr/lib/python3/dist-packages/nose/util.py", line 453, in try_run inspect.getargspec(func) File "/usr/lib/python3.5/inspect.py", line 1040, in getargspec stacklevel=2) File "/usr/lib/python3/dist-packages/numpy/testing/utils.py", line 2199, in _showwarning self._orig_show(message, category, filename, lineno, AttributeError: 'suppress_warnings' object has no attribute '_orig_show'
Bug#849177: Python-numpy transition breaks at least two packages
Hi all, On 23.12.2016 11:19, Andreas Tille wrote: > I'm also not sure and people who know better than me should feel free to > close bug #849177. I'd be really happy if there are better solutions to > fix python-skimage in the next 36 hours. My reading of the arguments in > #849177 is that there are other reasons to revert the upgrade even > without the additional problem - thus I was opening the bug report > independently from other packages like python-skimage and python-skbio. While the original FTBFS for skimage is solved now, I found another problem with numpy 1.12 RC (and skimage) which I reported upstream: https://github.com/numpy/numpy/issues/8413 Currently, this keeps up to get a version of skimage into Debian that has a chance to migrate. This is really very disappointing to run into those problems just a few hours before the window for the stretch migration closes. Sandro, could you please *immediately* rollback the numpy RC or at least upload a version that otherwise solves all the regressions that are still open? Please put a stable, working version of numpy that we all can rely on. Time matters here. Best regards Ole (quite angry in the moment, sorry.)
Bug#849196: Sometimes, supress_warnings misses one of its attributes
Just to show an extremal case: https://buildd.debian.org/status/fetch.php?pkg=skimage=all=0.12.3-3=1482501775 has this 15 times. This is a regression; it did not happen with 1.11. Please fix this regression ASAP so that skimage can migrate safely before the freeze. Best Ole
Bug#848782: statsmodels: FTBFS: TypeError: 'float' object cannot be interpreted as an index
OK, I will re-upload directly to ftp-master. Andreas, since you already moved and cleaned up skimage: would you do the same for pandas and statsmodels? For statsmodels, I will already put the new VCS into d/control and do a team upload (pushing the changes as soon as the git is there). Cheers Ole On 25.12.2016 14:42, Yaroslav Halchenko wrote: > On December 25, 2016 7:29:09 AM EST, Ole Streicher <oleb...@debian.org> wrote: >> Hi Yaroslav, Michael,Andreas, >> >> I uploaded the NMU as release 1.1 to DELAYED/2 now. I preferred NMU >> since I am not a NeuroDebian team member. >> >> And I would (again) propose to move the package to debian-science. It >> would also be useful to have it with the default structure (with >> upstream and pristine-tar branches). >> >> Best regards >> >> Ole >> >> On 25.12.2016 10:43, Ole Streicher wrote: >>> I'll upload to DELAYED ASAP (need to learn how to do this anyway); >>> however the unresolved problem is still the pandas FTBFS. Without >> that, >>> statsmodels will not migrate. >>> >>> Cheers >>> >>> Ole >>> >>> On 25.12.2016 08:34, Andreas Tille wrote: >>>> Hi Ole, >>>> >>>>> the attached patch fixes the FTBFS. What is the schedule to fix the >>>>> other problems with statsimage -- especially pandas? >>>> >>>> I learned that NeuroDebian is happy about team uploads / NMUs - so >>>> I think its fine if you upload your patch. >>>> >>>> Kind regards >>>> >>>> Andreas. >>>> >>>> PS: If you are not convinced and there is no response from Yaroslav >> or >>>> Michael in the next 24 hours - well, people are occupied by real >> life >>>> these days, yust tell me and I'd upload. >>>> > > You can go ahead and upload without delays. > If you see somehow that moving under another bigger team would make package > better, you have my blessing for pandas and statsmodels > > Cheers >
Bug#848782: statsmodels: FTBFS: TypeError: 'float' object cannot be interpreted as an index
Hi Yaroslav, Michael,Andreas, I uploaded the NMU as release 1.1 to DELAYED/2 now. I preferred NMU since I am not a NeuroDebian team member. And I would (again) propose to move the package to debian-science. It would also be useful to have it with the default structure (with upstream and pristine-tar branches). Best regards Ole On 25.12.2016 10:43, Ole Streicher wrote: > I'll upload to DELAYED ASAP (need to learn how to do this anyway); > however the unresolved problem is still the pandas FTBFS. Without that, > statsmodels will not migrate. > > Cheers > > Ole > > On 25.12.2016 08:34, Andreas Tille wrote: >> Hi Ole, >> >>> the attached patch fixes the FTBFS. What is the schedule to fix the >>> other problems with statsimage -- especially pandas? >> >> I learned that NeuroDebian is happy about team uploads / NMUs - so >> I think its fine if you upload your patch. >> >> Kind regards >> >> Andreas. >> >> PS: If you are not convinced and there is no response from Yaroslav or >> Michael in the next 24 hours - well, people are occupied by real life >> these days, yust tell me and I'd upload. >>
Bug#848782: statsmodels: FTBFS: TypeError: 'float' object cannot be interpreted as an index
I'll upload to DELAYED ASAP (need to learn how to do this anyway); however the unresolved problem is still the pandas FTBFS. Without that, statsmodels will not migrate. Cheers Ole On 25.12.2016 08:34, Andreas Tille wrote: > Hi Ole, > >> the attached patch fixes the FTBFS. What is the schedule to fix the >> other problems with statsimage -- especially pandas? > > I learned that NeuroDebian is happy about team uploads / NMUs - so > I think its fine if you upload your patch. > > Kind regards > > Andreas. > > PS: If you are not convinced and there is no response from Yaroslav or > Michael in the next 24 hours - well, people are occupied by real life > these days, yust tell me and I'd upload. >
Bug#848859: FTBFS randomly (failing tests)
Hi Santiago, > In particular, if something happens 1 every 20 times on average, the > fact that it did not happen when you try 10 times does not mean in any > way that it may not happen. I must however say that I don't see why a package that fails to build once in 20 builds would have a release critical problem: There is nothing in our policy that requires a reproducible or even a always successful build. I totally agree that catching random failures is a good quality measure, but this is IMO severity "important" at maximum. And it would be nice to have this QA test not just before the release, but already early in the release cycle. This would help a lot to avoid stressful bugfixes just before the freeze. Best regards Ole
Bug#848505: tycho2: FTBFS (segfault during build)
Control: reassign -1 astrometry.net Control: affects -1 src:tycho2 Control: tags -1 pending It is confirmed that the crash is a problem in astrometry.net. A patch is available; we just clarify whether a new astrometry.net upstream version is issued or the patch is applied to the current version.
Bug#848782: statsmodels: FTBFS: TypeError: 'float' object cannot be interpreted as an index
Control: tags -1 patch Hi, the attached patch fixes the FTBFS. What is the schedule to fix the other problems with statsimage -- especially pandas? best regards Ole >From 5c6f9ffecf8e53c32edc14326042bb4d5ca6a641 Mon Sep 17 00:00:00 2001 From: Ole Streicher <oleb...@debian.org> Date: Sat, 24 Dec 2016 14:56:07 +0100 Subject: [PATCH] Fix index type in `reshape` to be integer --- debian/changelog | 6 ++ debian/patches/fix_wrong_index_type.patch | 15 +++ debian/patches/series | 1 + 3 files changed, 22 insertions(+) create mode 100644 debian/patches/fix_wrong_index_type.patch diff --git a/debian/changelog b/debian/changelog index febebbe..bbd71aa 100644 --- a/debian/changelog +++ b/debian/changelog @@ -1,3 +1,9 @@ +statsmodels (0.8.0~rc1+git59-gef47cd9-2) UNRELEASED; urgency=medium + + * Fix index type in `reshape` to be integer. Closes: #848782 + + -- Ole Streicher <oleb...@debian.org> Sat, 24 Dec 2016 14:56:22 +0100 + statsmodels (0.8.0~rc1+git59-gef47cd9-1) unstable; urgency=medium * Fresh upstream rc snapshot which hopefully addresses some diff --git a/debian/patches/fix_wrong_index_type.patch b/debian/patches/fix_wrong_index_type.patch new file mode 100644 index 000..ca284b9 --- /dev/null +++ b/debian/patches/fix_wrong_index_type.patch @@ -0,0 +1,15 @@ +Author: Ole Streicher <oleb...@debian.org> +Descrption: Fix index type in `reshape` to be integer. + +This will avoid FTBFS with numpy >= 1.12 beta +--- a/statsmodels/genmod/generalized_estimating_equations.py b/statsmodels/genmod/generalized_estimating_equations.py +@@ -2392,7 +2392,7 @@ + exog_names = [x.split("[")[0] for x in exog_names] + + params = np.reshape(self.params, +-(ncut, len(self.params) / ncut)) ++(ncut, len(self.params) // ncut)) + + for ev in exog_values: + diff --git a/debian/patches/series b/debian/patches/series index 0f4a162..649de9c 100644 --- a/debian/patches/series +++ b/debian/patches/series @@ -5,3 +5,4 @@ up_reduce_test_precision deb_skip_test_ons390 deb_no_nbformat_for_now up_3239.patch +fix_wrong_index_type.patch -- 2.10.2
Bug#849271: version comparison fails with numpy 1.12.0b1
package: astropy version: 1.3-1 severity: serious forwarded: https://github.com/astropy/astropy/issues/5643 With astropy 1.3 and numpy 1.12.0b1, many affiliated packages now fail in the tests when the numpy version number is compared. For example aplpy: Traceback (most recent call last): File "debian/tests/python3-aplpy", line 7, in import aplpy File "/usr/lib/python3/dist-packages/aplpy/__init__.py", line 9, in from ._astropy_init import * File "/usr/lib/python3/dist-packages/aplpy/_astropy_init.py", line 116, in from astropy import config File "/usr/lib/python3/dist-packages/astropy/__init__.py", line 116, in _check_numpy() File "/usr/lib/python3/dist-packages/astropy/__init__.py", line 104, in _check_numpy from .utils import minversion File "/usr/lib/python3/dist-packages/astropy/utils/__init__.py", line 20, in from .compat.odict import OrderedDict File "/usr/lib/python3/dist-packages/astropy/utils/compat/__init__.py", line 14, in from .numpycompat import * File "/usr/lib/python3/dist-packages/astropy/utils/compat/numpycompat.py", line 24, in NUMPY_LT_1_12 = not minversion('numpy', '1.12dev') File "/usr/lib/python3/dist-packages/astropy/utils/introspection.py", line 159, in minversion return parse_version(have_version) >= parse_version(version) File "/usr/lib/python3.5/distutils/version.py", line 70, in __ge__ c = self._cmp(other) File "/usr/lib/python3.5/distutils/version.py", line 337, in _cmp if self.version < other.version: TypeError: unorderable types: int() < str()
Bug#846002: blends-tasks must not be priority:important - ballot proposal
Hi Margarita, On 26.12.2016 14:43, Margarita Manterola wrote: > While it's great that there is a possibly better solution in the works for the > tasksel screen, there's still the issue at hand that blends-tasks has used the > severity of the package as a way of circumventing collaboration with tasksel > maintainers. I already several times pointed that out: We did not "circumvent" the collaboration with the tasksel maintainers. The relevant bug #758116 was assigned to tasksel for quite a long time, and the proposed solution to create a package blends-tasks was announced exactly there -- so the proposal was on the desk of the tasksel maintainers, explicitly asking if they would veto. The first response from a tasksel maintainer came only five weeks later, without a veto, but with a discussion that finally lead to the current wording and selection. This is not what I would usually call "circumvent collaboration". For any other critics, the usual way to collaborate is to open a bug report. This did not happen then before of this bug (#846002). #846002 was again based on the first version of blends-tasks (see Holgers initial message), and he failed to rebase it within an reasonable time (and the original critics were already handled by the update to blends-tasks 0.6.94 as described above). And, again, IMO the d-i team several times expressed their heavy overload. It seems natural that working on the blends selection in tasksel or the correct wording should be done by the people who actually deal with the blends, avoiding to put additional load on them. Could you support your accusation a bit? I have the feeling that you are ignoring my arguments (since I have put them here already). Best regards Ole
Bug#848758: Latest upgrade of Numpy breaks tests of other packages (Was: Bug#848758: python-skbio: FTBFS: Test failures)
Hi Adrian, On 21.12.2016 15:29, Adrian Bunk wrote: > On Wed, Dec 21, 2016 at 02:05:57PM +0100, Ole Streicher wrote: >> ... >> The according functions are since then marked as "deprecated" and issue >> a warning. Numpy introduced this change with 1.11rc already, which was >> uploaded to Debian about 6 months ago. After it appears that many >> packages (als in Debian) have problems with this, they postponed it for >> another release -- which happens just now. >> ... > > You are saying you knew for 6 months that many packages in Debian > have problems with the changes that have just been uploaded. > > Why were no bugs filed a few months before the upload? Most of my own programs have CI tests which immediately gave a signal. I forwarded this upstream (mainly as github issues) and took part in the diskussion between upstream (f.e. astropy) and numpy, so that finally the decision was made to postpone. Since I looked only on my own packages, there was no need to file bugs; especially since the problem was postponed. The upstream maintainers that were informed took measures in between; therefore I have only one package (astroml) currently failing. I had no information about any other package; therefore I couldn't file bug reports. I can just recommend: If a Python package comes with tests, use the Debian CI infrastructure for it. This helps a lot as an early indicator of problems. And since the problems appear on packages with build time test, I would like to ask back: why don't you use CI? Best regards Ole
Bug#846094: cpl_mask.c: Unaligned access
Control: tags -1 pending Hi, I applied the patch from Ubuntu in the git repository. However, since the bug is severity "normal" and we are in freeze, I do not upload a new revision. Ubuntu patch: https://patches.ubuntu.com/c/cpl/cpl_7.0-3ubuntu1.patch Relevant commit: https://anonscm.debian.org/cgit/debian-astro/packages/cpl.git/commit/?id=1654c4f5d72d768f40a9f1415697049e7704282f Best regards Ole
Bug#858260: pandas: FTBFS (pandas.tests.test_multilevel.TestMultiLevel fails)
Looking into ci.debian.net, the new failure appears for the first time on 2017-03-14: https://ci.debian.net/data/packages/unstable/amd64/p/pandas/20170314_082050.autopkgtest.log.gz The debci log also shows an update of tzdata for this time: https://ci.debian.net/data/packages/unstable/amd64/p/pandas/20170314_082050.log so I would suspect this as the cause as well.
Bug#857067: Excluding arch s390x for dsdp?
Control: tags -1 patch Hi Andreas and all, it seems that this bug is fixed in Ubuntu; at least there is a quite careful analysis: https://bugs.launchpad.net/ubuntu/+source/dsdp/+bug/1543982 I attach the complete patch; it needs some minor clean-up, but is a good candidate to solve this. Cheers Ole diff -pruN 5.8-9.1/debian/changelog 5.8-9.1ubuntu2/debian/changelog --- 5.8-9.1/debian/changelog 2012-05-27 20:01:57.0 + +++ 5.8-9.1ubuntu2/debian/changelog 2016-04-14 11:52:55.0 + @@ -1,3 +1,21 @@ +dsdp (5.8-9.1ubuntu2) xenial; urgency=medium + + * Cast INFO to int before storing it in the flag. LP: #1543982. + + -- Dimitri John LedkovThu, 14 Apr 2016 12:52:28 +0100 + +dsdp (5.8-9.1ubuntu1) xenial; urgency=medium + + * Build using -O2 on s390x. LP: #1543982. + + -- Matthias Klose Wed, 10 Feb 2016 11:28:13 +0100 + +dsdp (5.8-9.1build1) xenial; urgency=medium + + * No-change rebuild for libblas3gf->libblas3 transition. + + -- Matthias Klose Sat, 16 Jan 2016 16:13:00 +0100 + dsdp (5.8-9.1) unstable; urgency=low * Non-maintainer upload. diff -pruN 5.8-9.1/debian/control 5.8-9.1ubuntu2/debian/control --- 5.8-9.1/debian/control 2011-07-22 07:29:20.0 + +++ 5.8-9.1ubuntu2/debian/control 2016-04-14 11:53:55.0 + @@ -1,7 +1,8 @@ Source: dsdp Section: science Priority: extra -Maintainer: Soeren Sonnenburg +Maintainer: Ubuntu Developers +XSBC-Original-Maintainer: Soeren Sonnenburg Build-Depends: cdbs, debhelper (>= 5), gfortran, libblas-dev, liblapack-dev, doxygen, doxygen-latex Standards-Version: 3.9.2 diff -pruN 5.8-9.1/debian/patches/series 5.8-9.1ubuntu2/debian/patches/series --- 5.8-9.1/debian/patches/series 1970-01-01 00:00:00.0 + +++ 5.8-9.1ubuntu2/debian/patches/series 2016-04-14 11:53:35.0 + @@ -0,0 +1 @@ +type-mismatch.patch diff -pruN 5.8-9.1/debian/patches/type-mismatch.patch 5.8-9.1ubuntu2/debian/patches/type-mismatch.patch --- 5.8-9.1/debian/patches/type-mismatch.patch 1970-01-01 00:00:00.0 + +++ 5.8-9.1ubuntu2/debian/patches/type-mismatch.patch 2016-04-14 11:53:49.0 + @@ -0,0 +1,15 @@ +Description: Cast INFO to int before storing it in the flag +Author: Dimitri John Ledkov +Bug-Ubuntu: https://bugs.launchpad.net/bugs/1543982 + +--- dsdp-5.8.orig/src/vecmat/dlpack.c dsdp-5.8/src/vecmat/dlpack.c +@@ -123,7 +123,7 @@ static int DTPUMatCholeskyFactor(void* A + dtpuscalemat(AP,ss,N); + } + dpptrf(, , AP, ); +- *flag=INFO; ++ *flag=(int) INFO; + return 0; + } + diff -pruN 5.8-9.1/debian/rules 5.8-9.1ubuntu2/debian/rules --- 5.8-9.1/debian/rules 2012-05-27 19:07:13.0 + +++ 5.8-9.1ubuntu2/debian/rules 2016-04-14 11:53:19.0 + @@ -5,15 +5,18 @@ DEB_SHLIBDEPS_INCLUDE := /usr/lib/atlas ver:= $(DEB_UPSTREAM_VERSION) soname := libdsdp-$(ver)gf.so +DEB_HOST_ARCH := $(shell dpkg-architecture -qDEB_HOST_ARCH) +OPT = -O3 + common-build-arch:: debian/build-arch-stamp debian/build-arch-stamp: - $(MAKE) DSDPROOT=`pwd` OPTFLAGS="-fPIC -O3" dsdpapi LAPACKBLAS="-llapack \ + $(MAKE) DSDPROOT=`pwd` OPTFLAGS="-fPIC $(OPT)" dsdpapi LAPACKBLAS="-llapack \ -lblas -lm" cd lib && ar x libdsdp.a && \ $(CC) -shared -Wl,--no-add-needed,-soname=$(soname) -o $(soname) *.o \ -llapack -lblas -lm && ln -s $(soname) libdsdp.so && rm *.o $(MAKE) DSDPROOT=`pwd` -C examples clean - $(MAKE) DSDPROOT=`pwd` OPTFLAGS="-O3" DSDPLIB="-L$(CURDIR)/lib -ldsdp -lm" \ + $(MAKE) DSDPROOT=`pwd` OPTFLAGS="$(OPT)" DSDPLIB="-L$(CURDIR)/lib -ldsdp -lm" \ example LAPACKBLAS="" touch $@ diff -pruN 5.8-9.1/.pc/applied-patches 5.8-9.1ubuntu2/.pc/applied-patches --- 5.8-9.1/.pc/applied-patches 1970-01-01 00:00:00.0 + +++ 5.8-9.1ubuntu2/.pc/applied-patches 2016-04-15 01:37:08.336449243 + @@ -0,0 +1 @@ +type-mismatch.patch diff -pruN 5.8-9.1/.pc/type-mismatch.patch/src/vecmat/dlpack.c 5.8-9.1ubuntu2/.pc/type-mismatch.patch/src/vecmat/dlpack.c --- 5.8-9.1/.pc/type-mismatch.patch/src/vecmat/dlpack.c 1970-01-01 00:00:00.0 + +++ 5.8-9.1ubuntu2/.pc/type-mismatch.patch/src/vecmat/dlpack.c 2005-10-21 19:31:15.0 + @@ -0,0 +1,1015 @@ +#include "dsdpsys.h" +#include "dsdpvec.h" +#include "dsdplapack.h" + +/*! \file dlpack.c +\brief DSDPDataMat, DSDPDualMat, DSDPDSMat, DSDPSchurMat, DSDPXMat, +objects implemented in dense upper packed symmetric format. +*/ + +typedef struct{ + char UPLO; + double *val; + double *v2; + double *sscale; + int scaleit; + int n; + int owndata; +} dtpumat; + +static const char* lapackname="DENSE,SYMMETRIC,PACKED STORAGE"; + +int DTPUMatEigs(void*AA,double W[],double IIWORK[], int nn1, double *mineig){ + dtpumat* AAA=(dtpumat*) AA; + ffinteger info,INFO=0,M,N=AAA->n; + ffinteger IL=1,IU=1,LDZ=1,IFAIL; + ffinteger *IWORK=(ffinteger*)IIWORK; + double