On 21/02/20 12:38, Tobias Burnus wrote:
Hmm, that sounds like papering over a real bug.
It is possible, I tried using DECL_INITIAL to nullify cfi.n but it did not made any difference.
I tried to play with optimization and up to -O1 it does not seem to crash but it always seems to crash at -O2 (without the -static-libgfortran switch). Looking at the generated code I could not see anything obviously different...
Before the patch to PR92123 there were situations where there would be off bounds memory writes, but that would cause a crash at a later time.
I have no idea of what it could be... Best regards, José Rui
