[PATCH v5 00/68] Define _GNU_SOURCE for sources using

Edward Liaw edliaw at google.com
Wed May 22 10:44:36 PDT 2024


On Wed, May 22, 2024 at 10:13 AM Jakub Kicinski <kuba at kernel.org> wrote:
>
> On Wed, 22 May 2024 10:19:33 -0600 Shuah Khan wrote:
> > On 5/21/24 18:56, Edward Liaw wrote:
> > > Centralizes the definition of _GNU_SOURCE into KHDR_INCLUDES and removes
> > > redefinitions of _GNU_SOURCE from source code.
> > >
> > > 809216233555 ("selftests/harness: remove use of LINE_MAX") introduced
> > > asprintf into kselftest_harness.h, which is a GNU extension and needs
> >
> > Easier solution to define LINE_MAX locally. In gerenal it is advisable
> > to not add local defines, but it is desirable in some cases to avoid
> > churn like this one.
>
> Will the patch that Andrew applied:
> https://lore.kernel.org/all/20240519213733.2AE81C32781@smtp.kernel.org/
> make its way to Linus? As you say that's a much simpler fix.

Right, this patch series may be unnecessary after all, since the
problem is fixed by that patch.

It might be better to drop the series unless it is desirable to
centralize the declaration of _GNU_SOURCE to the root Makefile /
lib.mk.  If that is still wanted, maybe a more palatable approach
would be to surround every instance of #define _GNU_SOURCE with
#ifndef _GNU_SOURCE first, then induce the change to CFLAGS in lib.mk.
That would prevent a partial merge from triggering build warnings.



More information about the linux-riscv mailing list