Bug#904319: FTBFS when building binary-packages only
Hi, On 07/23/2018 05:58 PM, Graham Inggs wrote: > dpkg-buildpackage -b works fine for me. it does if the upstream tarball is present in the parent directory, which is not a requirement when building a binary package from the source tree, no? Regards, Daniel
Bug#904319: FTBFS when building binary-packages only
Control: severity -1 normal Control: tags -1 + moreinfo unreproducible Hi Daniel On 23/07/2018 11:54, Daniel Baumann wrote: However, when locally rebuilding the package I saw that it wants to check the md5sum of the upstream tarball. This fails when building binary packages only, which should be possible. dpkg-buildpackage -b works fine for me. How did you obtain the source package and what error message did you see? Regards Graham
Bug#904319: FTBFS when building binary-packages only
Thanks for the full log, very helpful! I can reproduce the failure with LC_ALL=POSIX python3 ... and it could be avoided by LC_ALL=C.UTF-8 python3 ... On Mon, Jul 23, 2018 at 02:43:41PM +0200, Santiago Vila wrote: > On Mon, Jul 23, 2018 at 12:16:45PM +, Lumin wrote: > > > Thanks for the report, but could you please provider a more verbose > > failure report? > > Sure. I've put the full build log here: > > https://people.debian.org/~sanvila/build-logs/intel-mkl/ > > This was a binary-indep only build, i.e. "dpkg-buildpackage -A". > > Assuming "dpkg-buildpackage -A" is not the issue here, you would get full > build logs here as well: > > https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/intel-mkl.html > > but only when the page finally exists. > > > And could you please test this patch: > > > > LC_ALL=C.UTF-8 python3 debian/control.py > > Sorry, I can't test modified packages easily in my autobuilder setup. > > You should be able to reproduce this using either sbuild or pbuilder. > By looking at the proposed patch I bet that this could be reproduced > with a simple chroot as well, setting LC_ALL=C before dpkg-buildpackage, > but I don't really know. > > Thanks.
Bug#904319: FTBFS when building binary-packages only
On Mon, Jul 23, 2018 at 12:16:45PM +, Lumin wrote: > Thanks for the report, but could you please provider a more verbose > failure report? Sure. I've put the full build log here: https://people.debian.org/~sanvila/build-logs/intel-mkl/ This was a binary-indep only build, i.e. "dpkg-buildpackage -A". Assuming "dpkg-buildpackage -A" is not the issue here, you would get full build logs here as well: https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/intel-mkl.html but only when the page finally exists. > And could you please test this patch: > > LC_ALL=C.UTF-8 python3 debian/control.py Sorry, I can't test modified packages easily in my autobuilder setup. You should be able to reproduce this using either sbuild or pbuilder. By looking at the proposed patch I bet that this could be reproduced with a simple chroot as well, setting LC_ALL=C before dpkg-buildpackage, but I don't really know. Thanks.
Bug#904319: FTBFS when building binary-packages only
Hi Santiago, Thanks for the report, but could you please provider a more verbose failure report? And could you please test this patch: LC_ALL=C.UTF-8 python3 debian/control.py ``` diff --git a/debian/rules b/debian/rules index 212c9aa..8d21adc 100755 --- a/debian/rules +++ b/debian/rules @@ -39,7 +39,7 @@ autogen: extract-rpms $(AUTOGEN_FILES) chmod +x debian/libmkl-dev.postinst debian/libmkl-dev.prerm debian/libmkl-dev.config override_dh_auto_configure: autogen - python3 debian/control.py # Generate install files and lintian overrides + LC_ALL=C.UTF-8 python3 debian/control.py # Generate install files and lintian overrides # deal with embedded libjs-jquery $(RM) opt/intel/documentation_2018/ja/mkl/ps2018/resources/jquery-1.11.1.min.js ```
Bug#904319: FTBFS when building binary-packages only
severity 904319 serious reassign 904319 src:intel-mkl found 904319 2018.3.222-1 thanks On Mon, 23 Jul 2018, Daniel Baumann wrote: > Package: intel-mkl > Version: 2018.3.222-1 > > Hi, > > thank you so much for packaging intel-mkl, it basically doubles the > performance of our R number crunchers. > > However, when locally rebuilding the package I saw that it wants to > check the md5sum of the upstream tarball. This fails when building > binary packages only, which should be possible. I get this error when building with "dpkg-buildpackage -A": UnicodeDecodeError: 'ascii' codec can't decode byte 0xc2 in position 2262: ordinal not in range(128) make[1]: *** [debian/rules:42: override_dh_auto_configure] Error 1 make[1]: Leaving directory '/<>' make: *** [debian/rules:23: build-indep] Error 2 dpkg-buildpackage: error: debian/rules build-indep subprocess returned exit status 2 Dear maintainer: Now that the package has been accepted in unstable, try uploading in source only form (dpkg-buildpackage -S). This way bugs like this will never propagate to testing. Thanks.
Bug#904319: FTBFS when building binary-packages only
Package: intel-mkl Version: 2018.3.222-1 Hi, thank you so much for packaging intel-mkl, it basically doubles the performance of our R number crunchers. However, when locally rebuilding the package I saw that it wants to check the md5sum of the upstream tarball. This fails when building binary packages only, which should be possible. Regards, Daniel