Simon Tournier <zimon.toutoune@gmail.com> writes:

> Clicking on the “3” from <https://qa.guix.gnu.org/issue/61620> leads to
> <https://data.qa.guix.gnu.org/compare/package-derivations?base_commit=3809aa01a1287dbe5be90e7aa4407bf27a6467ed&target_commit=e2112b0d1e90a5a1f4f82bce8f8a29ab5fd86408&system=x86_64-linux&target=none&build_change=broken>
> which displays nothing:
>
>     Showing 0 new derivations across 0 packages
>
> Moreover, clicking on
> <https://data.qa.guix.gnu.org/compare/package-derivations?base_commit=3809aa01a1287dbe5be90e7aa4407bf27a6467ed&target_commit=e2112b0d1e90a5a1f4f82bce8f8a29ab5fd86408&system=x86_64-linux&target=none>
> lists many that does not seem related to ’tlpui’.
>
> This story looks like #61616.
>
> <http://issues.guix.gnu.org/msgid/782ad1100907d6b1487103ab37faedf52abc9019.camel@gmail.com>
>
> Since in both cases it seems an issue with Open MPI and it had been
> fixed by 9ae4846c502b75867b83180bb65d422ff838e4e6, are these 3 failures
> relevant?

They are indeed not relevant to the changes here. They are detected as
changing because the derivations do change, but that's because they
change every time you try and compute them, rather than changing as a
result from the patches here.

I think a way to avoid this would be to have the error happen earlier,
so rather than waiting until you try and build the derivation, try to
detect that the builder script can't be read when it's being created. I
can't see an easy way to do that, but I've filed an issue about it:
https://issues.guix.gnu.org/62051