On 07/10/17 15:03, Joerg Sonnenberger wrote:
On Mon, Jul 10, 2017 at 02:16:22PM -0500, Jason Bacon wrote:On 07/10/17 13:45, Joerg Sonnenberger wrote:On Mon, Jul 10, 2017 at 10:30:28AM -0500, Jason Bacon wrote:New issue as of today, on CentOS 6:Please start by filling a bug report with your distribution. warn_unused_result on most stdio.h function is utterly bogus as they provide persistent means of checking for error conditions. Whoever adds the flags is just breaking existing code for absolutely no good reason. JoergFYI... https://gnats.netbsd.org/cgi-bin/query-pr-single.pl?number=52390CentOS, not NetBSD. The bug is in the broken glibc headers. Joerg
Well, I can simply disable PKGSRC_USE_FORTIFY during bootstrap, but I'd prefer a solution that leaves security measures intact.
I think it's reasonable for an optional security feature like FORTIFY_SOURCE to throw a warning for ignoring the return value of an I/O function that might fail and leave a corrupted output file, so I would not say the headers are broken.
The fwrite() calls in fixup-libtool.c may not be likely to fail, but my inclination would be to either silence these warnings by adding checks to the code, or make the warnings non-fatal to the pkg build by default.
Thanks,
JB
--
Earth is a beta site.