https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82089
--- Comment #6 from Tom de Vries <vries at gcc dot gnu.org> --- (In reply to Andrew Stubbs from comment #5) > Any reason not to get this committed? No, this should get committed. This should either: - be fixed as part of upstreaming the gcn port, or - we can trigger the problem on bfin using the afore-mentioned trigger patch, and test that in combination with the fix, and prove that there are no regressions compared a vanilla bfin build and test, and see it this level of testing is acceptable for a middle-end maintainer.
