Add extra builds to test for reproducible builds

[bmwiedemann]

This helps slow down introduction of new issues into Factory.

See https://reproducible-builds.org/ for why this is good.

This needs added prjconf in a central place:

%if %_repository == "rb_future1y"
VMinstall: reproducible-faketools-futurepost
%endif

%if %_repository == "rb_j1"
VMinstall: reproducible-faketools-j1
BuildFlags: nochecks
%endif

BuildFlags: nodisturl
Release: 1.1

Macros:
%distribution reproducible
:Macros

The nochecks part is optional.

[DimStar77]

%if %_repository == "rb_j1" tends to not work across project boundaries, a <path> element makes prjconf from the parent being included (good) - but it resets %_repository the the repository name of the <path> element, thus not having the effect you expect here.

This needs to be replicated to all staging projects (incl. the dynamically setup adi projects)

e.g. osc buildconfig openSUSE:Factory:Staging:M standard | grep "%_repository"

[DimStar77]

how are the results from rb_future1y and rb_j1 compared to be rebuildable? They merely seem to test that 'they build in the future and with -j1, but there seems to be no test so far to ensure they 'are the same' ?

[bmwiedemann]

I pushed an extra commit to add prjconf in the script.

[dirkmueller]

so that's testing with -j1 ? how much slowdown is that causing?

[bmwiedemann]

https://rb.zq1.de/compare.factory/reproducible.json has duration values for -j1 and -j4 .
For most smaller packages, it does not make much of a difference, because non-build times make up for the majority of time.
For MozillaFirefox, it says

      "build_duration"  : 39489,
      "build_duration4" : 22782,

and for bash

      "build_duration"  : 363,
      "build_duration4" : 274,

However, nothing needs to block on this build to finish.

[dirkmueller]

what would be the most descriptive name? maybe standard_1yinfuture ?

[DimStar77]

As discussed this morning, the reproducible state should be used for informational purpose (to start with) and not be blocking. Later on we can probably let the bot accept on confirmed or ask a human for confirmation when not confirmed.

To not block when any of the reproducible builds fail (e.g timebomb in > 1 year) we discussed moving the reproducible builds out of the 'staging dashboard observed areas;

e.g openSUSE:Factory:Staging:adi:65:reproducible (redoing what adi:65 does)

This meta and prjconf snippets worked in my experiments:

prj meta:

<project name="openSUSE:Factory:Staging:adi:65:reproducible">
  <title/>
  <description/>
  <link project="openSUSE:Factory:Staging:adi:65"/>
  <url>/staging_workflows/openSUSE:Factory/staging_projects/openSUSE:Factory:Staging:adi:65</url>
  <person userid="dimstar" role="maintainer"/>
  <group groupid="factory-staging" role="maintainer"/>
  <publish>
    <disable/>
  </publish>
  <debuginfo>
    <enable/>
  </debuginfo>
  <repository name="rb_j1" linkedbuild="all">
    <path project="openSUSE:Factory:Staging" repository="standard"/>
    <path project="openSUSE:Factory" repository="standard"/>
    <arch>x86_64</arch>
  </repository>
  <repository name="rb_future1y" linkedbuild="all">
    <path project="openSUSE:Factory:Staging" repository="standard"/>
    <path project="openSUSE:Factory" repository="standard"/>
    <arch>x86_64</arch>
  </repository>
</project>

prjconf

%if "%_repository" == "rb_future1y"
Required: reproducible-faketools-futurepost
%endif

%if "%_repository" == "rb_j1"
Required: reproducible-faketools-j1
BuildFlags: nochecks
%endif

BuildFlags: nodisturl
Release: 1.1

Macros:
%distribution reproducible