Re: make dist broken
2012/2/15 Graham Percival gra...@percival-music.ca: Look, the make dist broken is just a subject line that I typed. I can't remember the exact message the GUB gave, but it's the usual one that happens whenever people add a new translation. Search the mailing list archives to see me complaining about it in the past, as well as explaining how to fix it. And/or look for such messages in the git commit log. Something like fix make dist or build: dist fix should probably find the right commits. Trust me, I really want to learn how to prevent dist problems caused by translations. Please enlighten me: is it necessary for me to run GUB so that I can detect dist failures? Given that my translation is a complete set (and not the only one which is) I can compare my Makefiles with those of other languages whenever a new file comes in, but I don't know if this is enough. Thanks. -- Francisco Vila. Badajoz (Spain) www.paconet.org , www.csmbadajoz.com ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
Francisco Vila paconet@gmail.com writes: 2012/2/15 Graham Percival gra...@percival-music.ca: Look, the make dist broken is just a subject line that I typed. I can't remember the exact message the GUB gave, but it's the usual one that happens whenever people add a new translation. Search the mailing list archives to see me complaining about it in the past, as well as explaining how to fix it. And/or look for such messages in the git commit log. Something like fix make dist or build: dist fix should probably find the right commits. Trust me, I really want to learn how to prevent dist problems caused by translations. Please enlighten me: is it necessary for me to run GUB so that I can detect dist failures? Given that my translation is a complete set (and not the only one which is) I can compare my Makefiles with those of other languages whenever a new file comes in, but I don't know if this is enough. I suppose we are all trying our best. And of course it is frustrating if you don't have a reliable method to make sure that your best is good enough for not disrupting the work of others. LilyPond is a complex project, and few of the people setting up complex parts of it like GUB, the build system, the translation infrastructure, remain in active duty. It would be nice to say we are working on it but I don't even know that. Basically we are trying to get on with the diluted knowledge we have in a civil manner (and it is not like my track record for that is exactly shining). With regard to translations, you are probably one of the persons best qualified for helping to find solutions to detect breakage from translations earlier than previously. The staging patchy does a full doc build by now, and that has helped. But it is not the whole deal, apparently. -- David Kastrup ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
Quoting Francisco Vila paconet@gmail.com: 2012/2/15 Graham Percival gra...@percival-music.ca: Look, the make dist broken is just a subject line that I typed. I can't remember the exact message the GUB gave, but it's the usual one that happens whenever people add a new translation. Search the mailing list archives to see me complaining about it in the past, as well as explaining how to fix it. And/or look for such messages in the git commit log. Something like fix make dist or build: dist fix should probably find the right commits. Trust me, I really want to learn how to prevent dist problems caused by translations. Please enlighten me: is it necessary for me to run GUB so that I can detect dist failures? Given that my translation is a complete set (and not the only one which is) I can compare my Makefiles with those of other languages whenever a new file comes in, but I don't know if this is enough. Thanks. -- Francisco Vila. Badajoz (Spain) www.paconet.org , www.csmbadajoz.com Replying from my webmail (working today) so the formatting may be odd and some of the explanation may be scratchy. The story of this one is that I noted error messages in make doc along the lines of Documentation/cs/included is not a directory. That was accurate - it didn't exist, and so I agreed with GP the simplest way to fix the error was to make the directory exist and put a file in it to ensure that it remained in git. I added a GNUmakefile to the included directories for cs, hu and ZH, IIRC. AFAICS this solved the problems. However, I believe Graham then kicked off a GUB build which creates a tarball of the source using make dist, and then checks that the contents of the tarball are the same as the source directories with make distcheck. They weren't, and this was what Graham was pointing out. The new GNUmakefiles and directories were in the source tree, but not the tarball. My research showed that this was fixed by updating the GNUmakefile in the Documentation/language. If you look at those, you'll see something like included_directories = list of dirs. I added included to cs, hu and zh and believe this fixed the problem. In practice, my attempt to push this to master using patchy failed, but it succeeded when Graham tried - so I don't understand that yet. make dist works on my normal build system, but I think make distcheck only works on GUB, so at present it's not easy to check for this problem. I think the message for translators (and me in this case) is that the makefiles do have lists of directories that need to be kept in step with the actual directories. I reckon this is a candidate for an issue - the directory list could be automatically created. This message was sent using IMP, the Internet Messaging Program. ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
On Thu, Feb 16, 2012 at 10:24:46AM +0100, Francisco Vila wrote: 2012/2/15 Graham Percival gra...@percival-music.ca: Look, the make dist broken is just a subject line that I typed. Trust me, I really want to learn how to prevent dist problems caused by translations. Great! Download GUB and look at test-lily/dist-check.py That's the script that complains. Of course, you'll need to find out what arguments to call that script with (see GUB's makefiles), and it might need the same directory layout that GUB uses... but that's where to look. Please enlighten me: is it necessary for me to run GUB so that I can detect dist failures? In theory? no. In practice? at the moment, yes. Extracting the dist-check from GUB and integrating it into the normal lilypond build system and scripts would likely take somebody like Julien or David 1 hour, it would probably be about 2 hours for me, and maybe 10 hours for you? no insult implied here; I'm just trying to guess how much you know about python and build systems. But I don't think anybody should be particularly concerned about this. There's at least 20 maintainability issues which have worse consequences for developers or potential contributors. Given that my translation is a complete set (and not the only one which is) I can compare my Makefiles with those of other languages whenever a new file comes in, but I don't know if this is enough. If you do a completely accuracy comparison, that should be enough. In other words, whenever there's been a problem, I can see differences in the GNUmakefiles. - Graham ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
- Original Message - From: Graham Percival gra...@percival-music.ca To: Phil Holmes m...@philholmes.net Cc: lilypond-devel@gnu.org Sent: Wednesday, February 15, 2012 10:41 PM Subject: Re: make dist broken On Wed, Feb 15, 2012 at 10:18:02PM -, Phil Holmes wrote: OK -as remarked in my previous mail patchy-merge fails with: I currently have no idea why this is. ok, I'll try running it. - Graham FWIW it was segfaulting on one of the midi regression tests. Couldn't see a reason for this, so logged out of patchy and back in again and all was well (I tend to leave patchy and my development username logged in and hibernate the box). -- Phil Holmes ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
make dist broken
file from VC not distributed: lilypond-2.15.30/Documentation/cs/included/GNUmakefile file from VC not distributed: lilypond-2.15.30/Documentation/hu/included/GNUmakefile file from VC not distributed: lilypond-2.15.30/Documentation/zh/included/GNUmakefile - Graham ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
On Wed, Feb 15, 2012 at 07:03:36PM -, Phil Holmes wrote: Well - that's clearly down to my adding those files to .git. What does make dist do? It, or rather make distcheck, makes sure that the source tarball matches the actual source. But I don't know exactly what GUB calls, so don't bother looking for the precise make distcheck. Instead, look at how the other languages are handled in Documentation/GNUmakefile, and do the same thing for your three new languages. Or maybe look at Documentation/??/GNUmakefile. One of those would be it. - Graham ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
- Original Message - From: Graham Percival gra...@percival-music.ca To: Phil Holmes m...@philholmes.net Cc: lilypond-devel@gnu.org Sent: Wednesday, February 15, 2012 7:08 PM Subject: Re: make dist broken On Wed, Feb 15, 2012 at 07:03:36PM -, Phil Holmes wrote: Well - that's clearly down to my adding those files to .git. What does make dist do? It, or rather make distcheck, makes sure that the source tarball matches the actual source. But I don't know exactly what GUB calls, so don't bother looking for the precise make distcheck. Instead, look at how the other languages are handled in Documentation/GNUmakefile, and do the same thing for your three new languages. Or maybe look at Documentation/??/GNUmakefile. One of those would be it. - Graham So - presumably make dist creates the tarball, and as part of it, it calls make distcheck to compare the tarball with the source tree. Does that sound sensible? -- Phil Holmes ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
- Original Message - From: Graham Percival gra...@percival-music.ca To: lilypond-devel@gnu.org Cc: Phil Holmes em...@philholmes.net Sent: Wednesday, February 15, 2012 6:50 PM Subject: make dist broken file from VC not distributed: lilypond-2.15.30/Documentation/cs/included/GNUmakefile file from VC not distributed: lilypond-2.15.30/Documentation/hu/included/GNUmakefile file from VC not distributed: lilypond-2.15.30/Documentation/zh/included/GNUmakefile - Graham Well - that's clearly down to my adding those files to .git. What does make dist do? -- Phil Holmes ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
On Wed, Feb 15, 2012 at 07:14:53PM -, Phil Holmes wrote: So - presumably make dist creates the tarball, and as part of it, it calls make distcheck to compare the tarball with the source tree. Does that sound sensible? In normal unix packages, 'make dist' createa a tarball, then 'make distcheck' expands the tarball, runs configure make make clean, then compares the result. Something like that. Look, the make dist broken is just a subject line that I typed. I can't remember the exact message the GUB gave, but it's the usual one that happens whenever people add a new translation. Search the mailing list archives to see me complaining about it in the past, as well as explaining how to fix it. And/or look for such messages in the git commit log. Something like fix make dist or build: dist fix should probably find the right commits. - Graham ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
- Original Message - From: Graham Percival gra...@percival-music.ca To: Phil Holmes m...@philholmes.net Cc: lilypond-devel@gnu.org Sent: Wednesday, February 15, 2012 7:20 PM Subject: Re: make dist broken On Wed, Feb 15, 2012 at 07:14:53PM -, Phil Holmes wrote: So - presumably make dist creates the tarball, and as part of it, it calls make distcheck to compare the tarball with the source tree. Does that sound sensible? In normal unix packages, 'make dist' createa a tarball, then 'make distcheck' expands the tarball, runs configure make make clean, then compares the result. Something like that. Look, the make dist broken is just a subject line that I typed. I can't remember the exact message the GUB gave, but it's the usual one that happens whenever people add a new translation. Search the mailing list archives to see me complaining about it in the past, as well as explaining how to fix it. And/or look for such messages in the git commit log. Something like fix make dist or build: dist fix should probably find the right commits. - Graham I did look, but in the end it was easier to find the problem by looking at what happened in make dist. Documentation/lang/GNUmakefile has a line like: SUBDIRS = essay learning notation texidocs usage web extending included which lists the subdirectories. I'm 99% certain that adding include to the new languages will fix the problem. You want me to do this, test and push to staging then master via patchy? FWIW I can't _really_ check, because my system doesn't have a rule for distcheck, but I can check the tarball from make dist. -- Phil Holmes ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
On Wed, Feb 15, 2012 at 08:24:23PM -, Phil Holmes wrote: I did look, but in the end it was easier to find the problem by looking at what happened in make dist. Documentation/lang/GNUmakefile has a line like: SUBDIRS = essay learning notation texidocs usage web extending included which lists the subdirectories. I'm 99% certain that adding include to the new languages will fix the problem. So am I. You want me to do this, test and push to staging then master via patchy? Yes. FWIW I can't _really_ check, because my system doesn't have a rule for distcheck, but I can check the tarball from make dist. Don't bother checking. Just stick it in staging and run staging-merge. - Graham ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
- Original Message - From: Graham Percival gra...@percival-music.ca To: Phil Holmes m...@philholmes.net Cc: lilypond-devel@gnu.org Sent: Wednesday, February 15, 2012 8:55 PM Subject: Re: make dist broken On Wed, Feb 15, 2012 at 08:24:23PM -, Phil Holmes wrote: I did look, but in the end it was easier to find the problem by looking at what happened in make dist. Documentation/lang/GNUmakefile has a line like: SUBDIRS = essay learning notation texidocs usage web extending included which lists the subdirectories. I'm 99% certain that adding include to the new languages will fix the problem. So am I. You want me to do this, test and push to staging then master via patchy? Yes. FWIW I can't _really_ check, because my system doesn't have a rule for distcheck, but I can check the tarball from make dist. Don't bother checking. Just stick it in staging and run staging-merge. - Graham I checked with a make dist, and the files are in the tarball. Pushed to staging and patchy is running. -- Phil Holmes ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
- Original Message - From: Phil Holmes m...@philholmes.net To: Graham Percival gra...@percival-music.ca Cc: lilypond-devel@gnu.org Sent: Wednesday, February 15, 2012 9:04 PM Subject: Re: make dist broken - Original Message - From: Graham Percival gra...@percival-music.ca To: Phil Holmes m...@philholmes.net Cc: lilypond-devel@gnu.org Sent: Wednesday, February 15, 2012 8:55 PM Subject: Re: make dist broken On Wed, Feb 15, 2012 at 08:24:23PM -, Phil Holmes wrote: I did look, but in the end it was easier to find the problem by looking at what happened in make dist. Documentation/lang/GNUmakefile has a line like: SUBDIRS = essay learning notation texidocs usage web extending included which lists the subdirectories. I'm 99% certain that adding include to the new languages will fix the problem. So am I. You want me to do this, test and push to staging then master via patchy? Yes. FWIW I can't _really_ check, because my system doesn't have a rule for distcheck, but I can check the tarball from make dist. Don't bother checking. Just stick it in staging and run staging-merge. - Graham I checked with a make dist, and the files are in the tarball. Pushed to staging and patchy is running. Hmm. Patchy fails. Looking at it. -- Phil Holmes ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
- Original Message - From: Phil Holmes m...@philholmes.net To: Graham Percival gra...@percival-music.ca Cc: lilypond-devel@gnu.org Sent: Wednesday, February 15, 2012 9:04 PM Subject: Re: make dist broken - Original Message - From: Graham Percival gra...@percival-music.ca To: Phil Holmes m...@philholmes.net Cc: lilypond-devel@gnu.org Sent: Wednesday, February 15, 2012 8:55 PM Subject: Re: make dist broken On Wed, Feb 15, 2012 at 08:24:23PM -, Phil Holmes wrote: I did look, but in the end it was easier to find the problem by looking at what happened in make dist. Documentation/lang/GNUmakefile has a line like: SUBDIRS = essay learning notation texidocs usage web extending included which lists the subdirectories. I'm 99% certain that adding include to the new languages will fix the problem. So am I. You want me to do this, test and push to staging then master via patchy? Yes. FWIW I can't _really_ check, because my system doesn't have a rule for distcheck, but I can check the tarball from make dist. Don't bother checking. Just stick it in staging and run staging-merge. - Graham I checked with a make dist, and the files are in the tarball. Pushed to staging and patchy is running. -- Phil Holmes OK -as remarked in my previous mail patchy-merge fails with: Success: nice make test -j9 CPU_COUNT=9 *** FAILED BUILD *** nice make doc -j9 CPU_COUNT=9 Previous good commit: bf7bb9a675ece83b468c84b499b1546db0f995e6 Current broken commit: ff5127b57881478bc27ea3321e473c5d0d78d983 I currently have no idea why this is. I've pulled staging to a clean distro and make doc runs fine. It's too late for me to do anything tonight and I'm working tomorrow, so it'll be tomorrow evening until I can contribute again. If anyone can see what's happening in the meantime, I'd appreciate it. For me, I'll next concentrate on better diagnostics from patchy-staging - I can't see why it failed. -- Phil Holmes ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: make dist broken
On Wed, Feb 15, 2012 at 10:18:02PM -, Phil Holmes wrote: OK -as remarked in my previous mail patchy-merge fails with: I currently have no idea why this is. ok, I'll try running it. - Graham ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel