RE: [r17-603 Regression] FAIL: g++.dg/opt/shrink-wrapping-vector-1.C -std=gnu++98 scan-rtl-dump pro_and_epilogue "Performing shrink-wrapping" on Linux/x86_64

> From: Andrew Pinski <[email protected]>
> Sent: Thursday, May 21, 2026 9:58 AM
> 
> On Tue, May 19, 2026 at 10:09 PM Haochen Jiang
> <[email protected]> wrote:
> >
> > On Linux/x86_64,
> >
> > acfdad706d8acca6c4ed6ef3b63c2e02f1c47881 is the first bad commit
> > commit acfdad706d8acca6c4ed6ef3b63c2e02f1c47881
> > Author: Nathan Myers <[email protected]>
> > Date:   Fri Apr 3 23:55:39 2026 -0400
> >
> >     libstdc++: Use allocate_at_least in vector, string (P0401) [PR118030]
> >
> > caused
> >
> > FAIL: g++.dg/opt/shrink-wrapping-vector-1.C  -std=gnu++98  scan-rtl-dump
> pro_and_epilogue "Performing shrink-wrapping"
> 
> I filed https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125400 for the
> missed optimization which is causing the missed shrink wrapping.
> If I can figure out how to xfail it for C++98 only I will unless
> someone beats me to it. Note It fails on all the specified targets in
> the testcase as it is generic issue.
> 
 
The interesting thing is that I don't know why, even before this commit, this test
seems to be failing under my env. I suppose my env could corrupt for one commit
but should not corrupt under each commit build.

XFAIL is a good idea since this is supposed to fail.

Thx,
Haochen