Bug#1093412: Regression in tests with new faketime

[James Addison]

On Sun, 19 Jan 2025 at 19:31, Ian Jackson
<ijackson at chiark.greenend.org.uk> wrote:
>
> Ian Jackson writes ("Re: Bug#1093412: Regression in tests with new faketime"):
> > Why are we passing NO_FAKE_STAT here?
> >
> > Maybe the problem is that faketime isn't honouring NO_FAKE_STAT.
> > I will see if I can confirm...
>
> (I should say, thanks for the lead!)

You're welcome; thanks for reading into it.  Here's my approximate
understanding:

Although we want reprotest to intentionally vary[1] the apparent
system clock -- to help discover unwanted dynamic timestamps that
appear in the resulting output -- we do not want it to affect
filesystem timestamps, because those may represent authorship and/or
versioning information.  In some cases I think filesystem timestamps
can also affect build processes - e.g. Make target evaluation - but
I'm not sure that that is our primary concern.

By default, faketime does adjust apparent filesystem timestamps -- and
so we use the NO_FAKE_STAT option[2] to deactivate that behaviour.

[1] - https://manpages.debian.org/bookworm/reprotest/reprotest.1.en.html#Time

[2] - https://manpages.debian.org/bookworm/faketime/faketime.1.en.html#ADVANCED_USAGE