[RFC 0/6] Fix lost siginfo_t

[RFC 0/6] Fix lost siginfo_t

[Jan Kratochvil]

Hi,

currently GDB's postponing and resending signals across threads drops the
associated siginfo_t information.  This can break real world inferiors
	https://bugzilla.redhat.com/attachment.cgi?id=413842
and it is also tested by a new testcase.

IMO all current `int status', `int host_signal' and `enum target_signal'
should be associated with their siginfo_t as otherwise they carry incomplete
information.  This is implemented by this patchset although it brings
sizeof (target_signal) to 136.  It is currently passed by value everywhere, it
could be either just shortened to 48 bytes or some pointer / reference
counting could be implemented.  It is true I would not change it myself
without a request as x86* host machines cannot notice such a small overhead.

There are some details to tune up but I would welcome to get some preliminary
approval of the approach before some more details + ChangeLogs get tuned up.

No regressions on {x86_64,x86_64-m32,i686}-fedora13-linux-gnu, with new KFAIL:
http://sourceware.org/bugzilla/show_bug.cgi?id=11842
Currently it has some runtime + compilation regressions on non-x86* arches.


Thanks,
Jan

Re: [RFC 0/6] Fix lost siginfo_t

[Mark Kettenis]

> Date: Tue, 27 Jul 2010 00:50:54 +0200
> From: Jan Kratochvil <jan.kratochvil@redhat.com>
> 
> Hi,
> 
> currently GDB's postponing and resending signals across threads drops the
> associated siginfo_t information.  This can break real world inferiors
> 	https://bugzilla.redhat.com/attachment.cgi?id=413842
> and it is also tested by a new testcase.
> 
> IMO all current `int status', `int host_signal' and `enum
> target_signal' should be associated with their siginfo_t as
> otherwise they carry incomplete information.  This is implemented by
> this patchset although it brings sizeof (target_signal) to 136.  It
> is currently passed by value everywhere, it could be either just
> shortened to 48 bytes or some pointer / reference counting could be
> implemented.  It is true I would not change it myself without a
> request as x86* host machines cannot notice such a small overhead.

Sorry, but I think the idea here is seriously flawed.  How are you
going to pass around the siginfo_t on non-native/multi-arch GDB?

obsoleted: Re: [RFC 0/6] Fix lost siginfo_t

[Jan Kratochvil]

On Tue, 27 Jul 2010 10:01:21 +0200, Mark Kettenis wrote:
> Sorry, but I think the idea here is seriously flawed.  How are you
> going to pass around the siginfo_t on non-native/multi-arch GDB?

Thanks, addressed in the new patchset:
	[0/9]#2 Fix lost siginfo_t
	http://sourceware.org/ml/gdb-patches/2010-08/msg00480.html