Re: gdbserver/gdb-6/4 and lots of pthreads

[jbbachky at aim dot com]

> On Fri, Nov 17, 2006 at 01:46:56PM -0500, jbbachky@aim.com wrote:
> > Sending packet:
$QPassSignals:e;10;11;13;14;17;18;1a;1b;1c;1e;1f;20;21;24;25;4c;#af...Ack

> Ignore signal 0xe.  In the GDB remote protocol that is SIGALRM.
I had "handle SIGALRM nostop noprint" in my .gdbinit file  I had added 
that while
digging into this problem a while ago and had forgotten about it.  
Removing it
and re-running everything doesn't appear to have an effect.  In 
addition to
that, I handle SIGPWR, SIGUSR1, SIGUSR2, SUGXCPU SIGCONT and SIGSTOP 
with
"nostop noprint"

With the handle SIGALRM removed, I get a steady stream of:

Sending packet: $vCont;C0e:1bb;c#ea...Ack
Packet received: T0e01:8ff7fb80;40:0f6bf3a8;thread:1bb;

perhaps 100/second.  Also, I see my other pthreads waking up and being 
switched-to
200 times/second.  When not attached to gdb, these do not wake up at 
all.  And when
I create an event which should wake up one of them, the system thrashes 
between that
one and the main parent thread, and "remote debug" data ceases.

> > Sending packet: $vCont;C0e:1c2;c#bb...Ack
> > Packet received: T0e01:8ff7fb80;40:0f6bf3a8;thread:1c2;
>
> Resume thread with SIGLARM.  Thread stopped with SIGALRM.
> It shouldn't be doing this; you told it to ignore SIGALRM.
>
> I can't reproduce this behavior; gdbserver doesn't report the SIGALRM
> for me.  Are you using unmodified versions from CVS at this point?

This is the CVS version from Wednesday evening, with your patch 
contained here:
 http://sources.redhat.com/ml/gdb-patches/2006-11/msg00114.html

If that doesn't have all of the pieces I need, please let me know - a 
few days I would
have said that I have no experience with patches/patching.  And now I 
have some :-)

> --
> Daniel Jacobowitz
> CodeSourcery

________________________________________________________________________
Check Out the new free AIM(R) Mail -- 2 GB of storage and 
industry-leading spam and email virus protection.