Hi Sebastian! Huh, I thought I had replied to this email, but apparently I had not yet.
On 2026-05-12T23:49:25+0200, Martin Jambor <[email protected]> wrote: > On Tue, Apr 21 2026, Sebastian Galindo via Gcc wrote: >> Sorry for the delay in sharing details about bug 93226. I have been >> spending time testing different things related to debugging and reading >> through the source code to (honestly) understand the fix, while the fix >> itself is straightforward, tracing the root cause involves following a >> fairly complete path through the compilation and execution flow. >> >> I left a detailed comment on Bugzilla explaining my reasoning for why >> the fix works and why I believe no additional tests are needed. Thanks, I shall look into that, soon, and also reply to your follow-on question. >> I may >> have over-explained a bit, but I wanted the record to be clear and also >> to check if my understanding is accurate. No worries; appreciate any such "verboseness". As you say, it helps us/me check that you're on the right track. >> My question is about regression testing. I would like to run the >> regression tests following the process described in the GCC newbies >> guide >> (https://gcc-newbies-guide.readthedocs.io/en/latest/readying-a-patch.html#regression-testing-bootstrap-regstrapping), >> >> but I am a bit unsure about which branch to target. Since the bug is in >> OG15 (i.e., devel/omp/gcc-15, if I am understanding correctly), should I >> run the regression tests against that branch, or against trunk/master? Thanks, Martin, for responding: > I would always recommend following master fairly closely unless you have > a specific reason not to (and even then, I'd recommend simply rebasing > less often, not track a release branch). However, if the issue you are > fixing really only manifests itself on the the OG15 branch (which is > still being used for development), then you indeed have to compare the > regression test results with and without your patch on that branch. The latter is indeed the case here: this is a bug fix for an OpenACC feature only present on OG15 but not yet trunk, just finally exposed (per my understanding) by one of the PR93226 test cases. So, yes, this must be tested on OG15. Grüße Thomas
