This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Notice clock_* functions in librt.so for old GLIBC_2.18 builds?
- From: Roland McGrath <roland at hack dot frob dot com>
- To: "Carlos O'Donell" <carlos at redhat dot com>
- Cc: Brooks Moses <bmoses at google dot com>, Andreas Schwab <schwab at linux-m68k dot org>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Sat, 22 Feb 2014 01:16:30 -0800 (PST)
- Subject: Re: Notice clock_* functions in librt.so for old GLIBC_2.18 builds?
- Authentication-results: sourceware.org; auth=none
- References: <53070029 dot 90407 at redhat dot com> <877g8opzuv dot fsf at igel dot home> <87y514oio2 dot fsf at igel dot home> <5307CB8F dot 4080507 at google dot com> <5307D2DD dot 702 at redhat dot com> <20140221235459 dot CDB4F744B4 at topped-with-meat dot com> <5307F2C3 dot 3020303 at redhat dot com>
> On 02/21/2014 06:54 PM, Roland McGrath wrote:
> >> The question that remains is: How do we prevent this from happening
> >> again?
Also, someone should try turning on -Wundef in the build and see if it
generates any warnings. It should generate one for the case of the missing
Versions.def entry. (It's easy to test in any configuration--even a
momentarily hand-hacked one--with a shlib-versions third column of
GLIBC_2.17 or newer, since you can drop the GLIBC_2.17 line from
Versions.def and check nm rt/clock-compat.os for being empty as it should
be.)
Perhaps something annoying will come up. But I tend to think that if
-Wundef turns up any warnings (other than an intentionally-induced one for
abi-versions.h misgeneration, which one should verify it does), we should
just fix the uses and then put -Werror=undef into +gccwarn. Ever intending
to evaluate undefined tokens in cpp expressions just seems sloppy off hand.
Thanks,
Roland