What happened in the Reproducible Builds effort between Sunday November 13 and Saturday November 19 2016:
Elsewhere in Debian
dpkg1.18.14 has migrated to stretch.
Ximin worked on glibc reproducibility this week, catching some bugs in disorderfs, FUSE, as well as glibc itself.
- Webconverger was added to our list of projects working on reproducible builds..
Packages reviewed and fixed, and bugs filed
- #844491 filed against linux by Chris Lamb.
- #844992 filed against perlbrew by Chris Lamb.
- #844993 filed against qsynth by Chris Lamb.
- #845034 filed against initramfs-tools by Chris Lamb.
- #845055 filed against markdown by Chris Lamb.
- #844752 filed against perl by Niko Tyni.
- #845007 filed against qsampler by Reiner Herrmann.
- #845008 filed against qmidinet by Reiner Herrmann.
- #845010 filed against qxgedit by Reiner Herrmann.
Reviews of unreproducible packages
43 package reviews have been added, 4 have been updated and 12 have been removed in this week, adding to our knowledge about identified issues.
2 issue types have been updated:
4 issue types have been added:
Weekly QA work
During our reproducibility testing, some FTBFS bugs have been detected and reported by:
- Chris Lamb (26)
- Daniel Stender (1)
- Filip Pytloun (1)
- Lucas Nussbaum (28)
- Michael Biebl (1)
- Don’t make tests rely on Debian::Debhelper::Dh_Lib. (Chris Lamb)
- #844498 (“disorderfs: using it for building kills the host”)
debrebuild is new tool proposed by HW42 and josch (see #774415: “From srebuild sbuild-wrapper to debrebuild”).
debrepatch is a set of scripts that we’re currently developing to make it easier to track unapplied patches. We have a lot of those and we’re not always sure if they still work. The plan is to set up jobs to automatically apply old reproducibility patches to newer versions of packages and notify the right people if they don’t apply and/or no longer make the package reproducible.
is a component of debrepatch that applies debdiffs to Debian source packages.
In other words, it is to
patch(1) is to
diff(1). It is a
general tool that is not specific to Reproducible Builds. This week, Ximin Luo
worked on making it more “production-ready” and will soon submit it for
inclusion in devscripts.
Ximin Luo significantly improved reprotest, adding presets and auto-detection
of which preset to use. One can now run e.g.
reprotest auto . or
auto $pkg_$ver.dsc instead of the long command lines that were needed before.
He also made it easier to set up build dependencies inside the virtual server and made it possible to specify pre-build dependencies that reprotest itself needs to set up the variations. Previously one had to manually edit the virtual server to do that, which was not very usable to humans without an in-depth knowledge of the building process.
These changes will be tested some more and then released in the near future as
- Ximin Luo:
- Set uname to two explicitly different values
- Add a –no-diffoscope option. (Closes: #844512)
- Add “auto” presets and –testbed-init
- Add –testbed-pre and –testbed-init arguments
- Fix main script so that build commands starting with ENVVAR=val work after append_command()
- Dramatically expand documentation (many commits)
- An index of our usertagged bugs page was added by Holger after a Q+A session in Cambridge.
- Holger also setup two new i386 builders, build12+16, for >50% increased build performance. For this, we went from 18+17 cores on two 48GB machines to 10+10+9+9 cores on four 36GB ram machines… and from 16 to 24 builder jobs. Thanks to Profitbricks for providing us with all these ressources once more!
- h01ger also tried to enable disorderfs again, but hit #844498, which brought down the i386 builders, so he disabled it again. Next will be trying disorderfs on armhf or amd64, to see whether this bug also manifests there.
This week’s edition was written by Chris Lamb, Holger Levsen, Ximin Luo and reviewed by a bunch of Reproducible Builds folks on IRC.