This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Weaken -fstack-protector configure test to a compile test
On Wed, 21 Mar 2012, Roland McGrath wrote:
> But one thing that comes to mind is if the old check might also have been
> verifying that GCC was not generating some external library call that is
> missing.
The trouble with that is that on platforms where external library
references are involved, it's references to __stack_chk_* identifiers that
the newly built glibc will definitely provide - and running a link test is
testing the wrong, old glibc for whether those functions are present
(although since the relevant symbol versions are GLIBC_2.4, it's not that
likely that the previous libc will be old enough for that to cause a wrong
result if there is a previous libc at all).
> Also, the old test actually compiled a function whereas now it's an empty
> input file. So if there is some error/warning that only arises when the
> compiler actually has anything to do, we could be missing it.
I don't think that's an issue; these sorts of warnings for unsupported
options are given at the end of option processing, without regard to
whether there is anything for the compiler to do.
--
Joseph S. Myers
joseph@codesourcery.com