Hi! On Fri, 2024-12-13 at 11:22:59 +0100, [email protected] wrote: > We found a unit test in the gcc package that killed dpkg-buildpackage > (#1089007). The error message was non obvious, so this adds a warning in > case someone stumbles of over this again.
As discussed on IRC when this was sent, this looked to me like it would at least benefit from sbuild giving detailed information about dpkg-buildpackage exit condition, which Johannes Schauer Marin Rodrigues improved in sbuild commit 8bfae61a0826aa34ccdcc6ad48743e276716da4c. Then I mentioned whether dpkg-buildpackage could provide even more information, such as the PID and UID for the process emitting such terminating signal, which we could do via sigaction's SA_SIGINFO. One question though, was whether it makes sense for dpkg-buildpackage to try to self report, as it will be unable to do that for example for signals that cannot be caught. And in the end it looks like this is best done from the parent driving its invocation, and where the specific condition that triggered this problem seems extremely unusual (a child committing a parricide). So I think after having thought about this a bit more, it looks like this is better done elsewhere (which we have now). Thanks for the patch though! Regards, Guillem
