On 2022-06-01, Holger Levsen wrote:
> On Sun, May 29, 2022 at 08:26:18PM +0200, Samuel Thibault wrote:
>> Zhang Boyang, le dim. 29 mai 2022 14:10:35 +0800, a ecrit:
>> > I found Salsa CI reprotest on my repo fails when "FAKETIME variation:
>> > faketime = [balabala]" is decided. The relevant output is:
>> > 
>> > dpkg-source: error: cannot change timestamp for
>> > build-experiment-1/.pc/applied-patches: Invalid argument
>> I have seen reprotest randomly fail in the salsa CI for various packages
>> indeed.
>
> indeed, merging with "#961064: reprotest: should not default to vary time and 
> date"
> and raising the severity and tagging help because reprotest is (almost) 
> unmaintained
> upstream, and tagging newcomer as the fix should be really easy, so if you are
> looking for ways to start contributing to reproducible builds, please do! ;p 
> :)
>
> I'll certainly be happy to review, merge and upload.

Given that the overwhelming majority of reproducible builds issues are
time related, and regressions happen all the time, disabling this would
fail to catch many issues, for what I believe to be a small minority of
affected packages.  I have personally run thousands of builds using
reprotest, and only rarely have come across this issue.

It typically causes issues with packages that patch files at build time,
such as applying debian/patches/ ... but most definitely not all
patches.

The salsa-ci pipelines document how to add custom arguments to reprotest:

  
https://salsa.debian.org/salsa-ci-team/pipeline#adding-extra-arguments-to-reprotest

So adding to your configuration:

  variables:
    SALSA_CI_REPROTEST_ARGS: --vary=-time

Might be good to update their documentation to use the --vary arguments
and include -time,-build_path ... I can maybe propose a patch to the
salsa-ci team for that.


live well,
  vagrant

Attachment: signature.asc
Description: PGP signature

Reply via email to