https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68298
Jeffrey A. Law <law at redhat dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2016-01-19
Ever confirmed|0 |1
--- Comment #9 from Jeffrey A. Law <law at redhat dot com> ---
Thanks for the .s and .o files. They were very helpful.
So if I compile the provided .o into an executable it still doesn't fail.
However, I can throw it into the debugger and see that we're reading a[400], so
depending on the exact memory layout and mapped pages we may or may not get a
fault. Sadly valgrind doesn't flag this.
>From looking at the provided .s file, we have clearly mis-compiled fn3. The
guard for the read of a[400] tests %al, which is never set.
If I compile this with a trunk compiler, the guard is properly initialized and
the read of a[400] never occurs. Bisecting now to find out if it was fixed or
just gone latent.
