Bug#907352: please make texinfo generation reproducible
Nicholas, > > (Is there a line-item in the issues.yml Reproducible Builds notes.git > > repo?) > > At this time reproducible-builds/reproducible-notes/issues.yml does > not contain a hint. (If it does not, please feel free to go ahead and add one and then tag the offending packages. This helps us get an idea of which toolchain issues to best prioritise.) Best wishes, -- ,''`. : :' : Chris Lamb `. `'` la...@debian.org / chris-lamb.co.uk `-
Bug#907352: please make texinfo generation reproducible
Hi Dmitry! On Mon, Aug 27, 2018 at 10:42:20AM +0300, Dmitry Shachnev wrote: > Control: forwarded -1 https://github.com/sphinx-doc/sphinx/pull/5359 > > Hi Nicholas! > > On Sun, Aug 26, 2018 at 04:32:51PM -0400, Nicholas D Steeves wrote: > > Here is an example of the unreproducibility on DebCI: > > > > https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/elpy.html > > I have created an upstream pull request that should fix this issue: > https://github.com/sphinx-doc/sphinx/pull/5359 > > Any testing/feedback is appreciated. Wow, thank you for working on this :-) Yes, if you'd like I can set up a local copy of src:python-sphinx, import your patch, test it, and forward the patch to this bug (or create a PR). If I haven't done this in 48h, please ping me. Cheers, Nicholas signature.asc Description: PGP signature
Bug#907352: please make texinfo generation reproducible
On Mon, Aug 27, 2018 at 07:43:57AM +0100, Chris Lamb wrote: > severity 907352 wishlist > thanks > > > I've set the severity of this bug to important, because I believe > > this bug affects all packages that generate info pages using Sphinx. > > Whilst this might affect all packages I don't believe the underlying > issue warrants that original severity and/or should be worked on before > other issues. That's fair, although I would have though normal was more appropriate. > (Is there a line-item in the issues.yml Reproducible Builds notes.git > repo?) At this time reproducible-builds/reproducible-notes/issues.yml does not contain a hint. That is why I had to resort to asking for help after the diffoscope baffled me :-) Cheers, Nicholas signature.asc Description: PGP signature
Bug#907352: please make texinfo generation reproducible
Control: forwarded -1 https://github.com/sphinx-doc/sphinx/pull/5359 Hi Nicholas! On Sun, Aug 26, 2018 at 04:32:51PM -0400, Nicholas D Steeves wrote: > Here is an example of the unreproducibility on DebCI: > https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/elpy.html I have created an upstream pull request that should fix this issue: https://github.com/sphinx-doc/sphinx/pull/5359 Any testing/feedback is appreciated. -- Dmitry Shachnev signature.asc Description: PGP signature
Bug#907352: please make texinfo generation reproducible
severity 907352 wishlist thanks > I've set the severity of this bug to important, because I believe > this bug affects all packages that generate info pages using Sphinx. Whilst this might affect all packages I don't believe the underlying issue warrants that original severity and/or should be worked on before other issues. (Is there a line-item in the issues.yml Reproducible Builds notes.git repo?) Regards, -- ,''`. : :' : Chris Lamb `. `'` la...@debian.org / chris-lamb.co.uk `-
Bug#907352: please make texinfo generation reproducible
Package: python3-sphinx Version: 1.7.7-1 Severity: important Dear Sphinx maintainers, Thank you for maintaining python3-sphinx. I've set the severity of this bug to important, because I believe this bug affects all packages that generate info pages using Sphinx. Here is an example of the unreproducibility on DebCI: https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/elpy.html I'm using this in debian/rules: PYTHONPATH=. sphinx-build -N -btexinfo docs/ build/info makeinfo --no-split build/info/Elpy.texi -o build/info/elpy.info Stuart Prescott helped me identify the cause on #debian-python and mentioned that it also affects "python-bumps", and I've CCed him. In #debian-mentors Jakub Wilk mentioned that it's possible to workaround with something like: PYTHONHASHSEED=0 It follows that a temporary workaround using a date stamp from the most recent changelog entry as a hash seed might not be terrible, but it would be much better if the root cause was addressed. Cheers, Nicholas