Re: GDB threading/FP bug

Re: GDB threading/FP bug

[Andrew Cagney]

> Some better searching revealed this is a very old bug.  From last
> year:
> 
> http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&th=adfbc404c374e980&seekm=OF36816CDD.AE7D54AF-ONC1256B1F.005622DC%40icos.be&frame=off
> 
> I tried running GDB in GDB and I was able to get as far as seeing that
> the floats seemed to be pushed onto the stack correctly until the
> thread switch and all the ptrace stuff, but I wasn't totally sure if
> the bug I was seeing at that point was in the GDB'd GDB or was the
> same bug in the outter GDB.

[Mailing list changed to something with less spam :-)]

As you've discovered, there have been a number of very similar bugs that 
all involving corrupted FP and threads.  It was thought that fixes for 
all such bugs were incorporated into both 5.2.* 
(http://sources.redhat.com/gdb/download/) and the mainline 
(http://sources.redhat.com/gdb/current/).

Can you please check that your problem does occur in the latest 
release/mainline, and if it does, create a bug report using 
http://sources.redhat.com/gdb/bugs/.

enjoy,
Andrew

Re: GDB threading/FP bug

[Kevin Buettner]

On Aug 8,  1:34pm, Andrew Cagney wrote:

> > Some better searching revealed this is a very old bug.  From last
> > year:
> > 
> > http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&th=adfbc404c374e980&seekm=OF36816CDD.AE7D54AF-ONC1256B1F.005622DC%40icos.be&frame=off
> > 
> > I tried running GDB in GDB and I was able to get as far as seeing that
> > the floats seemed to be pushed onto the stack correctly until the
> > thread switch and all the ptrace stuff, but I wasn't totally sure if
> > the bug I was seeing at that point was in the GDB'd GDB or was the
> > same bug in the outter GDB.
> 
> [Mailing list changed to something with less spam :-)]
> 
> As you've discovered, there have been a number of very similar bugs that 
> all involving corrupted FP and threads.  It was thought that fixes for 
> all such bugs were incorporated into both 5.2.* 
> (http://sources.redhat.com/gdb/download/) and the mainline 
> (http://sources.redhat.com/gdb/current/).
> 
> Can you please check that your problem does occur in the latest 
> release/mainline, and if it does, create a bug report using 
> http://sources.redhat.com/gdb/bugs/.

Yes, I thought this bug was fixed too.

FWIW, I just did a build of gdb-5.2.1 on a Red Hat 7.3 machine (x86). 
I wasn't able to reproduce the bug using the test program provided by
Bennet.  (I've also tested the current development sources and, as
expected, don't see the bug there either.)

Kevin

Re: GDB threading/FP bug

[bennet brauer]

Well I'm unable to compile from the head of CVS.  I get the following error 
as it tries to compile "po"...  This error doesn't exist in 5.2.1.

file=./`echo fr | sed 's,.*/,,'`.gmo \
  && rm -f $file && PATH=../src:$PATH no -o $file fr.po
/bin/sh: no: command not found


I did however find an old, closed, GDB bug that indicated linux 2.2.19 did 
not have this bug.  And indeed, on a 2.2.19 box, GDB doesn't exhibit this 
bug with my test program.  Currently I'm working around the issue with a 
dummy floating point operation at the top of my program.


>From: Kevin Buettner <kevinb@redhat.com>
>To: Andrew Cagney <ac131313@ges.redhat.com>, bennet 
><bennetbrauer@hotmail.com>
>CC: gdb@sources.redhat.com
>Subject: Re: GDB threading/FP bug
>Date: Thu, 8 Aug 2002 11:36:59 -0700
>
>On Aug 8,  1:34pm, Andrew Cagney wrote:
>
> > > Some better searching revealed this is a very old bug.  From last
> > > year:
> > >
> > > 
>http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&th=adfbc404c374e980&seekm=OF36816CDD.AE7D54AF-ONC1256B1F.005622DC%40icos.be&frame=off
> > >
> > > I tried running GDB in GDB and I was able to get as far as seeing that
> > > the floats seemed to be pushed onto the stack correctly until the
> > > thread switch and all the ptrace stuff, but I wasn't totally sure if
> > > the bug I was seeing at that point was in the GDB'd GDB or was the
> > > same bug in the outter GDB.
> >
> > [Mailing list changed to something with less spam :-)]
> >
> > As you've discovered, there have been a number of very similar bugs that
> > all involving corrupted FP and threads.  It was thought that fixes for
> > all such bugs were incorporated into both 5.2.*
> > (http://sources.redhat.com/gdb/download/) and the mainline
> > (http://sources.redhat.com/gdb/current/).
> >
> > Can you please check that your problem does occur in the latest
> > release/mainline, and if it does, create a bug report using
> > http://sources.redhat.com/gdb/bugs/.
>
>Yes, I thought this bug was fixed too.
>
>FWIW, I just did a build of gdb-5.2.1 on a Red Hat 7.3 machine (x86).
>I wasn't able to reproduce the bug using the test program provided by
>Bennet.  (I've also tested the current development sources and, as
>expected, don't see the bug there either.)
>
>Kevin




_________________________________________________________________
MSN Photos is the easiest way to share and print your photos: 
http://photos.msn.com/support/worldwide.aspx

Re: GDB threading/FP bug

[Daniel Jacobowitz]

On Tue, Aug 13, 2002 at 05:11:39PM -0300, bennet brauer wrote:
> Well I'm unable to compile from the head of CVS.  I get the following error 
> as it tries to compile "po"...  This error doesn't exist in 5.2.1.
> 
> file=./`echo fr | sed 's,.*/,,'`.gmo \
>  && rm -f $file && PATH=../src:$PATH no -o $file fr.po
> /bin/sh: no: command not found

Then you're missing a tool needed to compile GDB.   I think it's
gettext at that point.  Be sure you have that and libtool installed.

> 
> 
> I did however find an old, closed, GDB bug that indicated linux 2.2.19 did 
> not have this bug.  And indeed, on a 2.2.19 box, GDB doesn't exhibit this 
> bug with my test program.  Currently I'm working around the issue with a 
> dummy floating point operation at the top of my program.
> 
> 
> >From: Kevin Buettner <kevinb@redhat.com>
> >To: Andrew Cagney <ac131313@ges.redhat.com>, bennet 
> ><bennetbrauer@hotmail.com>
> >CC: gdb@sources.redhat.com
> >Subject: Re: GDB threading/FP bug
> >Date: Thu, 8 Aug 2002 11:36:59 -0700
> >
> >On Aug 8,  1:34pm, Andrew Cagney wrote:
> >
> >> > Some better searching revealed this is a very old bug.  From last
> >> > year:
> >> >
> >> > 
> >http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&th=adfbc404c374e980&seekm=OF36816CDD.AE7D54AF-ONC1256B1F.005622DC%40icos.be&frame=off
> >> >
> >> > I tried running GDB in GDB and I was able to get as far as seeing that
> >> > the floats seemed to be pushed onto the stack correctly until the
> >> > thread switch and all the ptrace stuff, but I wasn't totally sure if
> >> > the bug I was seeing at that point was in the GDB'd GDB or was the
> >> > same bug in the outter GDB.
> >>
> >> [Mailing list changed to something with less spam :-)]
> >>
> >> As you've discovered, there have been a number of very similar bugs that
> >> all involving corrupted FP and threads.  It was thought that fixes for
> >> all such bugs were incorporated into both 5.2.*
> >> (http://sources.redhat.com/gdb/download/) and the mainline
> >> (http://sources.redhat.com/gdb/current/).
> >>
> >> Can you please check that your problem does occur in the latest
> >> release/mainline, and if it does, create a bug report using
> >> http://sources.redhat.com/gdb/bugs/.
> >
> >Yes, I thought this bug was fixed too.
> >
> >FWIW, I just did a build of gdb-5.2.1 on a Red Hat 7.3 machine (x86).
> >I wasn't able to reproduce the bug using the test program provided by
> >Bennet.  (I've also tested the current development sources and, as
> >expected, don't see the bug there either.)
> >
> >Kevin
> 
> 
> 
> 
> _________________________________________________________________
> MSN Photos is the easiest way to share and print your photos: 
> http://photos.msn.com/support/worldwide.aspx
> 
> 

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer