This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFA] Remove REALTIME_HI/LO macros from target headers
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: drow at false dot org (Daniel Jacobowitz)
- Cc: gdb-patches at sourceware dot org
- Date: Tue, 28 Nov 2006 19:50:59 +0100 (CET)
- Subject: Re: [RFA] Remove REALTIME_HI/LO macros from target headers
Daniel Jacobowitz wrote:
> > For Linux, the nm-linux.h file already defines those macros, so the
> > definitions in the tm files appear to be fully redundant.
>
> The nm-linux.h define is redundant too, isn't it? And off by one?
> Compare to the top of signals.c. Or is the copy in signals.c off by
> one?
Right. The nm-linux.h define should be redundant too. However, it
is actually the correct version; the one at the top of signals.c
is off by one. (SIGRTMAX is the number of last real-time signal,
REALTIME_HI is supposed to be one beyond the number of the last
real-time signal.)
So we should fix the default definition in signals.c, and then
remove the redundant define in nm-linux.h. In fact, this would
allow to remove the #include <signal.h> from nm-linux.h as well,
but that *might* cause breakage if any of the native Linux targets
happens to rely on that implicit inclusion of <signal.h>.
That's unfortunately a bit hard to test for ...
> Patch looks OK modulo that question.
Would it be OK to commit the current patch as-is, and do the removal
from nm-linux.h as a follow-on patch?
Thanks,
Ulrich
--
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE
Ulrich.Weigand@de.ibm.com