This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [rfa, v2] Fix inferior calls during interrupted system calls on PowerPC
- From: Daniel Jacobowitz <drow at false dot org>
- To: Ulrich Weigand <uweigand at de dot ibm dot com>
- Cc: gdb-patches at sourceware dot org
- Date: Fri, 2 May 2008 10:36:03 -0400
- Subject: Re: [rfa, v2] Fix inferior calls during interrupted system calls on PowerPC
- References: <200803271933.m2RJXpRS017283@d12av02.megacenter.de.ibm.com> <200804251425.m3PEPGp6011396@d12av02.megacenter.de.ibm.com>
On Fri, Apr 25, 2008 at 04:25:10PM +0200, Ulrich Weigand wrote:
> Hello,
>
> this is an updated version of the patch to fix inferior system
> calls during interrupted system calls on PowerPC Linux:
> http://sourceware.org/ml/gdb-patches/2008-03/msg00452.html
>
> Retested on powerpc-linux and powerpc64-linux, both natively and
> using a local gdbserver. Fixes all interrupt.exp failures.
>
> Dan, are the remote/gdbserver parts of this patch OK?
You wrote:
> gdbserver is switched to always use powerpc-linux32/64 instead
> of powerpc-32/64. This still leaves the new registers out on
> any non-AltiVec platforms. I guess for powerpc-e500 we could
> create a similar powerpc-linuxe500 feature, if this is required.
> I'm not sure how to handle the default (non-AltiVec, non-SPE)
> case which does not yet use XML. Simply adding registers to
> the .dat files is probably not a good idea as it might break
> old GDBs that connect to the new gdbserver ...
I am not terribly concerned about breaking old GDBs. I think keeping
new GDB able to connect to old stubs is important, but if there's a
feature to add that means a new stub requires a new GDB that's usually
easy to accomodate - host programs are easier to upgrade than target
programs. We could switch the default case to use XML-generated
data files and add these.
I'd like to fix this on SPE too but there's clearly no obligation on
you to do that; next time we're working through our GDB test results
here at CodeSourcery we'll encounter the failure and hopefully notice
and fix it.
Unless there's a need to keep the Linux-specific register feature in
rs6000-tdep.c, it can go in ppc-linux-tdep.c; see mips-linux-tdep.c's
handling of the restart register for an example. Or is it more
complicated to get the numbering right that way?
--
Daniel Jacobowitz
CodeSourcery