[Reproducible-builds] GSoC 2015 Week 11: Move forward reproducible builds
Hi, This week I have updated my patch for gettext to target the new version uploaded to Debian unstable (0.19.5.1-1). I pushed the updated patch to our git repo: https://anonscm.debian.org/cgit/reproducible/gettext.git/commit/?h=pu/reproducible_builds&id=b6e972eab1f04f3fbd368433b496f1f4a6b372a3 I have also uploaded the new build to our custom APT repository, and attached the patch to the open bug: https://bugs.debian.org/792687 I have written a patch for the qthelpgenerator tool from qt4-x11 to honour SOURCE_DATE_EPOCH when embedding timestamps in the generated qch files (see issue timestamps_in_qch [1]). The tool qthelpgenerator embeds timestamps in different places, so in order to find a general solution I added the method reproducibleDateTime in the class QDateTime which returns a deterministic date honouring SOURCE_DATE_EPOCH in case the variable is set. Then I replaced the instances of QDateTime::currentDateTime() with QDateTime::reproducibleDateTime() where needed. I have pushed the patch to our git repo: https://anonscm.debian.org/cgit/reproducible/qt4-x11.git/commit/?h=pu/reproducible_builds&id=ee300f6185ce82eb932dbe1b2dca2113b88afb56 I have also uploaded the build to our custom APT repository, and opened a bug with the patch in Debian: https://bugs.debian.org/794681 And I have also updated the wiki accordingly. After patching qt4-x11, 3 packages became reproducible in my machine: - ktikz - qmf - linkchecker qmf didn't become reproducible in Jenkins: there is a weird issue with having different files in build 1 and 2 (apparently files are missing in both builds). This may be caused by parallelism issues and needs further investigation. There are still 12 packages that where tagged with timestamps_in_qch that didn't become reproducible. They are affected by randomness_in_qdoc [2]. Also some of them show variations in qch files (it seems to be a history of SQL commands that differ between builds). I have fixed some individual packages without notes: Affected by timestamps in zip metadata (timestamps_in_zip issue) - foxyproxy https://bugs.debian.org/794779 - xul-ext-monkeysphere https://bugs.debian.org/794781 Embedding of dates: - freeipmi https://bugs.debian.org/794792 - doc-base https://bugs.debian.org/794793 - debiandoc-sgml-doc https://bugs.debian.org/794795 For next week I'll probably look at the randomness_in_qdoc [2] issue. [1] https://reproducible.debian.net/issues/unstable/timestamps_in_qch_issue.html [2] https://reproducible.debian.net/issues/unstable/randomness_in_qdoc_issue.html Best regards, -- Dhole signature.asc Description: OpenPGP digital signature ___ Reproducible-builds mailing list Reproducible-builds@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds
[Reproducible-builds] GSoC 2015 Week 11: Move forward reproducible builds
Hello, This week I have been working on the following problems: 1. doxygen --- There was a new Debian doxygen release in unstable so I refreshed the version in the reproducible builds git repository and package repository. This took a bit longer because of the current libstdc++6 transition [1] which broke my system (I had to go use snapshot.d.o for a while) and because of two bugs I found in the process: - https://bugs.debian.org/794745 - https://bugs.debian.org/794758 2. camitk -- This package was wrongly tagged as suffering from timestamps_in_manpages_generated_by_doxygen [2]. The timestamps found in manpages were generated from a cmake script which has now been patched to use the environment variable $SOURCE_DATE_EPOCH [3] and packages.yml has been updated - https://bugs.debian.org/794740 3. gperf, maxima, and python-xlib -- These package were not affected by the latest texi2html release in unstable. When further investigated I found that they contained an embedded code copy of texi2html and they did not build-depend on the Debian texi2html. The copyright information of texi2html was missing from their debian/copyright. Package: gperf - https://bugs.debian.org/794955 Package: maxima - https://bugs.debian.org/795056 Package: python-xlib - https://bugs.debian.org/795057 4. texi2html - I modified the patch made by Juan Picca for texi2html to use $SOURCE_DATE_EPOCH when it produces its timestamps [4], I have also patched texi2html to sort its hash values, which solves the issue randomness_in_html_generated_by_texi2html [5]. The package has been QA uploaded to unstable with these fixes. - https://bugs.debian.org/783475 I scheduled 19 packages affected by texi2html. Out of these 2 became immediately reproducible, up till now 3 were found to not build-depend on texi2html but include an embedded code copy (see above) and the rest do not use dh or CDBS as their build system and therefore $SOURCE_DATE_EPOCH does not get exported and cannot be used by texi2html. Individual patches are needed for these packages. 5. Unfinished - I have been working on making bash reproducible and discovered that it produces unreproducible pdf through a mechanism that has not been inventoried yet. It uses dvipdfm. I will work on this issue more this week. I am also working on making the timestamps produced by pdftex be timezone independent. This will complete my initial pdftex patch to make the CreationDate, ModDate and ID field deterministic in pdf generated by latex [6] Kind Regards, akira [1] https://release.debian.org/transitions/html/libstdc++6.html [2] https://reproducible.debian.net/issues/unstable/timestamps_in_manpages_generated_by_doxygen_issue.html [3] https://wiki.debian.org/ReproducibleBuilds/TimestampsProposal [4] https://reproducible.debian.net/issues/unstable/timestamps_in_documentation_generated_by_texi2html_issue.html [5] https://wiki.debian.org/ReproducibleBuilds/RandomnessInHTMLGeneratedByTexi2html [6] https://reproducible.debian.net/issues/unstable/timestamps_in_pdf_generated_by_latex_issue.html signature.asc Description: OpenPGP digital signature ___ Reproducible-builds mailing list Reproducible-builds@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds
Re: [Reproducible-builds] GSoC 2015 Week 11: Move forward reproducible builds
Hi Dhole, On Samstag, 8. August 2015, Dhole wrote: > This week I have [...] (I think) you forgot to mention your blog post at https://dhole.github.io/post/reproducible_builds_debian_gsoc2015_update_1/ which I found really nice + very interesting to read, especially the two paragraphs "Impressions about reproducible builds" and "Impressions about working on a free software project"! :-) Thanks for all your work and reports! cheers, Holger signature.asc Description: This is a digitally signed message part. ___ Reproducible-builds mailing list Reproducible-builds@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds
Re: [Reproducible-builds] GSoC 2015 Week 11: Move forward reproducible builds
> This week I have been working on the following problems: [..] Great work :) Regards, -- ,''`. : :' : Chris Lamb `. `'` la...@debian.org / chris-lamb.co.uk `- ___ Reproducible-builds mailing list Reproducible-builds@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds
