Not a regression, but a patch to fix a bug that should have been ACK'd
long ago (as in years ago).
Basically Sergei proposed this patch back in July 2018:
https://gcc.gnu.org/legacy-ml/gcc-patches/2018-07/msg01791.html
Andreas objected to that patch. Sergei posted another which Rich
objected to.
Then in Sept there was another discussion where Andreas withdrew his
initial objection. I missed that and the patch sent into limbo state:
https://gcc.gnu.org/legacy-ml/gcc-patches/2018-09/msg00916.html
David Seifert reached out to me to get things resolved and after
reviewing the patch and the two threads apply the patch seemed like the
right thing to do. It looks technically correct, fixes a real bug and
while it's not a regression it can't have an effect on the code
generator as it just affects the runtime library (and just symbol
visibility in that). So I made an executive decision that it was OK to
go ahead and commit even though we're in stage4.
jeff
