I had the following conversation with richi about this patch.
Sorry to reply off thread, but i do net read this group in my mailer.
[09:00] zadeck richi: i am reviewing a patch and i have a couple
of questions, do you have a second to look at something?
[09:00] richi zadeck: sure
[09:01] zadeck the patch is
https://gcc.gnu.org/ml/gcc-patches/2015-07/msg01604.html
[09:01] zadeck he has set up the data flow problem correctly, but
i worry that this really is not the right way to solve this problem.
[09:02] richi let me look at the problem
[09:03] zadeck in particular, the only way he can demonstrate this
problem in c is with an uninitialized variable. It seems that a
normal correct program would not normally have this kind of issue unless
there was some upstream bug coming out of the middle end.
[09:04] richi even for his Ada case it's not initialized it seems
[09:06] richi zadeck: I've added a comment to the PR and requested
info
[09:06] zadeck my ada is as good as my German.
[09:07] zadeck the thing is that if you turn this into a truly
must problem, it will disable a lot of legit transformations.
[09:08] richi yep
[09:09] zadeck thanks
[09:09] richi lets see if he can point to a different issue
[09:09] richi or produce a C testcase for us
[09:13] jakub richi: yeah; if it is Ada only thing and Ada
uninitialized variable must have some special properties, then they'd
better use some on the side flag for whether it is initialized, or zero
initialize or whatever Ada expects
[09:15] richi jakub: from what I understand of Ada the testcase
doesn't look like such a case
[09:16] richi jakub: so the actual bug is likely somewhere else
Kenny