On 10/1/25 14:50, Aurelien Jarno wrote: >> What kind of testing was done on this ? > I have done a native build with --disable-bootstrap and ran the gcc > testsuite looking for regressions. I have then built glibc and ran the > its testsuite to confirm that the issue is fixed and that it doesn't > introduced any new issue. > >> You may wanna look at CI pre-commit trigger for this patch [1] >> It shows some lint fixes to address and more importantly shows additional >> ICE in >> fortran. > Thanks, I wasn't aware of this CI pre-commit trigger. And my local tests > predate this new unit test (building and testing takes time!).
Of course. FWIW in future if you submit any patches prefixed with RISC-V, they get tested automatically by Edwin's infra. And for non RISC-V patches (say middle-end) that you want tested for RISC-V, feel free to CC *[email protected] <mailto:[email protected]> * > Looking at the report, I have noticed that the patch indeed *introduced* > an ICE for gfortran.dg/pr87908.f90 for some targets and optimisation > levels, but at the same time it *fixed* the ICE for the same test and > for different targets and optimisation level. It looks like more a flaky > test than a real issue introduced by that bug. > > This seems to be confirmed by the fact this test has been removed in > commit 25f7f04e due to "memory leaks in f951". > > For the lint issues, I'll send a new version. Awesome ! Thx, -Vineet
