https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92364
--- Comment #3 from Gordon Mc <gmc at synopsys dot com> ---
Hi Andrew,
Thanks for your time, and for the quick response.
Yes, I may have reduced the testcase too far. It used to have the problem even
with the values being initialized, but that was a much larger testcase. I
don't still have the dozens of progressively reduced testcases that preceded
this one, so I may not be able to go back to that state without a lot of
rework.
I'm puzzled though, about how this issue is considered 'invalid', because it
seems to me that it clearly shows inconsistency within the compiler.
If the code is invalid, then surely it should be invalid regardless of whether
I build with "--coverage", and regardless of whether I disable inlining, etc.
I understand that the same piece of code may produce warnings / errors (or not
produce them) when I change compiler flags to enable or disable additional
checks, or change the language standard, etc., etc.
However, for simply enabling coverage or inlining - sure, that can result in
different paths being taken through the compiler, but I don't see how that can
justify changing the status from accepting the test code as valid / rejecting
the code. Can you possibly help clarify that for me, and / or point me at some
documentation that helps explain why the valid / invalid state of the code
should change based on things like coverage being enabled etc. ?
Thanks again for your help ; I realize the folks maintaining GCC may be unpaid
volunteers, so I definitely appreciate the time & effort you put into this.
Gordon
