PR 207, 662

PR 207, 662

[Smith, Jonathan C5]

GDB has been unusable on SGI IRIX 6.5 for some time now.  The only useful
information given by gdb is the following:

warning: Signal ? does not exist on this system.

But I know from attempts to debug that the hang is not immediately located
around that.  Not knowing how gdb works, I could only get so far in
attempting to debug this.  This hang occurs on any program I've attempted to
debug, not just the reported Python binary.

Is there anyone with any more information on this or some ability to solve
this?


Thank you,

Jonathan Smith
--
Jonathan C. Smith
Voice:  (301) 342-6367
Fax:     (301) 342-6381
Atlantic Ranges & Facilities
Computer Scientist
Building 2109 
48150 Shaw Road, Suite S115
Patuxent River, MD 20670-1907

 <<...OLE_Obj...>> 

Re: PR 207, 662

[Andrew Cagney]

Which GDB?  Have you tried a snap from the latest branch 
(ftp://sources.redhat.com/pub/gdb/snapshots/branch/)?

Andrew


> GDB has been unusable on SGI IRIX 6.5 for some time now.  The only useful
> information given by gdb is the following:
> 
> warning: Signal ? does not exist on this system.
> 
> But I know from attempts to debug that the hang is not immediately located
> around that.  Not knowing how gdb works, I could only get so far in
> attempting to debug this.  This hang occurs on any program I've attempted to
> debug, not just the reported Python binary.
> 
> Is there anyone with any more information on this or some ability to solve
> this?
> 
> 
> Thank you,
> 
> Jonathan Smith
> --
> Jonathan C. Smith
> Voice:  (301) 342-6367
> Fax:     (301) 342-6381
> Atlantic Ranges & Facilities
> Computer Scientist
> Building 2109 
> 48150 Shaw Road, Suite S115
> Patuxent River, MD 20670-1907
> 
>  <<...OLE_Obj...>> 
> 
> 

Re: PR 207, 662

[Andrew Cagney]

> Further, on attempt to compile, it fails in expect/exp_command.c with the
> following errors:
> 
> cc-1515 cc: ERROR File = exp_command.c, Line 920
>   A value of type "void *" cannot be assigned to an entity of type "int".
> 
>                 master = ((mode & TCL_READABLE)?rfd:wfd);
> 
> cc-1164 cc: ERROR File = exp_command.c, Line 930
>   Argument of type "void *" is incompatible with parameter of type "int".
> 
>                         if (-1 == (write_master = dup(wfd))) {

Hmm, I thought that one was `fixed'.  Just add a cast to supress the 
warning.

Andrew

Re: PR 207, 662

[Andrew Cagney]

> Using the same definitions of CC and CXX on the gdb nightly tarball dated
> today (20021118), the grep returned nothing (no results) with return value
> of 1.

Which suggests a problem in configure.  Does your system have one of the 
libraries I mentioned?

Beyond that, I'm afraid that you'll need to debug gdb/configure to see 
what the problem is.


Andrew

RE: PR 207, 662

[Smith, Jonathan C5]

Further, on attempt to compile, it fails in expect/exp_command.c with the
following errors:

cc-1515 cc: ERROR File = exp_command.c, Line 920
  A value of type "void *" cannot be assigned to an entity of type "int".

                master = ((mode & TCL_READABLE)?rfd:wfd);

cc-1164 cc: ERROR File = exp_command.c, Line 930
  Argument of type "void *" is incompatible with parameter of type "int".

                        if (-1 == (write_master = dup(wfd))) {

and various warnings.

Jonathan Smith

-----Original Message-----
From: Smith, Jonathan C5 
Sent: Monday, November 18, 2002 7:56 AM
To: 'Andrew Cagney'; Smith, Jonathan C5
Cc: 'gdb@sources.redhat.com'
Subject: RE: PR 207, 662


Using the same definitions of CC and CXX on the gdb nightly tarball dated
today (20021118), the grep returned nothing (no results) with return value
of 1.

Jonathan Smith

-----Original Message-----
From: Andrew Cagney [mailto:ac131313@redhat.com]
Sent: Friday, November 15, 2002 8:18 PM
To: Smith, Jonathan C5
Cc: 'gdb@sources.redhat.com'
Subject: Re: PR 207, 662


> On SGI 6.5:
>    It is still broken when built with GCC 3.2.
>    Setting the enviroment variable 'CC' to CC, the build breaks with
> 'configure: error: Could not find a term library' in '.../gdb'.

I just re-did a build on an IRIX 6.5 system using the 5.3 branch sources 
  I had:

     CC=/bin/cc
     CXX=/bin/CC

and configured with --disable-nls. Looking in gdb/config.status, 
configure found:

cagney@holodeck$ grep term config.status
s%@TERM_LIB@%-ltermlib%g
${ac_dA}HAVE_TERM_H${ac_dB}HAVE_TERM_H${ac_dC}1${ac_dD}
${ac_dA}HAVE_TERMIO_H${ac_dB}HAVE_TERMIO_H${ac_dC}1${ac_dD}
${ac_dA}HAVE_TERMIOS_H${ac_dB}HAVE_TERMIOS_H${ac_dC}1${ac_dD}

Is any of this on your system?

Andrew


> -----Original Message-----
> From: Andrew Cagney [mailto:ac131313@redhat.com]
> Sent: Wednesday, October 23, 2002 5:00 PM
> To: Smith, Jonathan C5
> Cc: 'gdb@sources.redhat.com'
> Subject: Re: PR 207, 662
> 
> 
> 
>> Snapshot from 20020403, 5.2.0, 5.2.1.
>> 
>> I haven't taken a snapshot from the 5.3 branch as it takes a few hours to
>> transfer to the stand-alone network.  I shall do that tonight.  Is there
a
>> reason to expect this fixed in the latest snapshot when its been a
problem
>> for a long time (PR 207 is dated Tue Sep 04 09:18:01 PDT 2001) and I
don't
>> see any resolution information in the PRs.
> 
> 
> A month or so back MichaelS KevinB and I gave the MIPS on IRIX stuff a 
> ``good talking too', n32, n64 and o32 now work a lot better.  It was all 
> on 6.5 with the IRIX compiler, from memory 207 refers to 5.x.
> 
> Andrew
> 
> 

RE: PR 207, 662

[Smith, Jonathan C5]

Using the same definitions of CC and CXX on the gdb nightly tarball dated
today (20021118), the grep returned nothing (no results) with return value
of 1.

Jonathan Smith

-----Original Message-----
From: Andrew Cagney [mailto:ac131313@redhat.com]
Sent: Friday, November 15, 2002 8:18 PM
To: Smith, Jonathan C5
Cc: 'gdb@sources.redhat.com'
Subject: Re: PR 207, 662


> On SGI 6.5:
>    It is still broken when built with GCC 3.2.
>    Setting the enviroment variable 'CC' to CC, the build breaks with
> 'configure: error: Could not find a term library' in '.../gdb'.

I just re-did a build on an IRIX 6.5 system using the 5.3 branch sources 
  I had:

     CC=/bin/cc
     CXX=/bin/CC

and configured with --disable-nls. Looking in gdb/config.status, 
configure found:

cagney@holodeck$ grep term config.status
s%@TERM_LIB@%-ltermlib%g
${ac_dA}HAVE_TERM_H${ac_dB}HAVE_TERM_H${ac_dC}1${ac_dD}
${ac_dA}HAVE_TERMIO_H${ac_dB}HAVE_TERMIO_H${ac_dC}1${ac_dD}
${ac_dA}HAVE_TERMIOS_H${ac_dB}HAVE_TERMIOS_H${ac_dC}1${ac_dD}

Is any of this on your system?

Andrew


> -----Original Message-----
> From: Andrew Cagney [mailto:ac131313@redhat.com]
> Sent: Wednesday, October 23, 2002 5:00 PM
> To: Smith, Jonathan C5
> Cc: 'gdb@sources.redhat.com'
> Subject: Re: PR 207, 662
> 
> 
> 
>> Snapshot from 20020403, 5.2.0, 5.2.1.
>> 
>> I haven't taken a snapshot from the 5.3 branch as it takes a few hours to
>> transfer to the stand-alone network.  I shall do that tonight.  Is there
a
>> reason to expect this fixed in the latest snapshot when its been a
problem
>> for a long time (PR 207 is dated Tue Sep 04 09:18:01 PDT 2001) and I
don't
>> see any resolution information in the PRs.
> 
> 
> A month or so back MichaelS KevinB and I gave the MIPS on IRIX stuff a 
> ``good talking too', n32, n64 and o32 now work a lot better.  It was all 
> on 6.5 with the IRIX compiler, from memory 207 refers to 5.x.
> 
> Andrew
> 
> 

Re: PR 207, 662

[Andrew Cagney]

> On SGI 6.5:
>    It is still broken when built with GCC 3.2.
>    Setting the enviroment variable 'CC' to CC, the build breaks with
> 'configure: error: Could not find a term library' in '.../gdb'.

I just re-did a build on an IRIX 6.5 system using the 5.3 branch sources 
  I had:

     CC=/bin/cc
     CXX=/bin/CC

and configured with --disable-nls. Looking in gdb/config.status, 
configure found:

cagney@holodeck$ grep term config.status
s%@TERM_LIB@%-ltermlib%g
${ac_dA}HAVE_TERM_H${ac_dB}HAVE_TERM_H${ac_dC}1${ac_dD}
${ac_dA}HAVE_TERMIO_H${ac_dB}HAVE_TERMIO_H${ac_dC}1${ac_dD}
${ac_dA}HAVE_TERMIOS_H${ac_dB}HAVE_TERMIOS_H${ac_dC}1${ac_dD}

Is any of this on your system?

Andrew


> -----Original Message-----
> From: Andrew Cagney [mailto:ac131313@redhat.com]
> Sent: Wednesday, October 23, 2002 5:00 PM
> To: Smith, Jonathan C5
> Cc: 'gdb@sources.redhat.com'
> Subject: Re: PR 207, 662
> 
> 
> 
>> Snapshot from 20020403, 5.2.0, 5.2.1.
>> 
>> I haven't taken a snapshot from the 5.3 branch as it takes a few hours to
>> transfer to the stand-alone network.  I shall do that tonight.  Is there a
>> reason to expect this fixed in the latest snapshot when its been a problem
>> for a long time (PR 207 is dated Tue Sep 04 09:18:01 PDT 2001) and I don't
>> see any resolution information in the PRs.
> 
> 
> A month or so back MichaelS KevinB and I gave the MIPS on IRIX stuff a 
> ``good talking too', n32, n64 and o32 now work a lot better.  It was all 
> on 6.5 with the IRIX compiler, from memory 207 refers to 5.x.
> 
> Andrew
> 
> 

RE: PR 207, 662

[Smith, Jonathan C5]

Further investigation has reduced this problem to binaries linked against
libpthread.  A simple cout "Hello world" should work when not linked against
libpthread, then fail when linked against it.

I hope this helps.....


Jonathan Smith

-----Original Message-----
From: Andrew Cagney [mailto:ac131313@redhat.com]
Sent: Wednesday, October 23, 2002 5:00 PM
To: Smith, Jonathan C5
Cc: 'gdb@sources.redhat.com'
Subject: Re: PR 207, 662


> Snapshot from 20020403, 5.2.0, 5.2.1.
> 
> I haven't taken a snapshot from the 5.3 branch as it takes a few hours to
> transfer to the stand-alone network.  I shall do that tonight.  Is there a
> reason to expect this fixed in the latest snapshot when its been a problem
> for a long time (PR 207 is dated Tue Sep 04 09:18:01 PDT 2001) and I don't
> see any resolution information in the PRs.

A month or so back MichaelS KevinB and I gave the MIPS on IRIX stuff a 
``good talking too', n32, n64 and o32 now work a lot better.  It was all 
on 6.5 with the IRIX compiler, from memory 207 refers to 5.x.

Andrew

RE: PR 207, 662

[Smith, Jonathan C5]

On SGI 6.5:
   It is still broken when built with GCC 3.2.
   Setting the enviroment variable 'CC' to CC, the build breaks with
'configure: error: Could not find a term library' in '.../gdb'.

Jonathan Smith

-----Original Message-----
From: Andrew Cagney [mailto:ac131313@redhat.com]
Sent: Wednesday, October 23, 2002 5:00 PM
To: Smith, Jonathan C5
Cc: 'gdb@sources.redhat.com'
Subject: Re: PR 207, 662


> Snapshot from 20020403, 5.2.0, 5.2.1.
> 
> I haven't taken a snapshot from the 5.3 branch as it takes a few hours to
> transfer to the stand-alone network.  I shall do that tonight.  Is there a
> reason to expect this fixed in the latest snapshot when its been a problem
> for a long time (PR 207 is dated Tue Sep 04 09:18:01 PDT 2001) and I don't
> see any resolution information in the PRs.

A month or so back MichaelS KevinB and I gave the MIPS on IRIX stuff a 
``good talking too', n32, n64 and o32 now work a lot better.  It was all 
on 6.5 with the IRIX compiler, from memory 207 refers to 5.x.

Andrew

Re: PR 207, 662

[Andrew Cagney]

> Snapshot from 20020403, 5.2.0, 5.2.1.
> 
> I haven't taken a snapshot from the 5.3 branch as it takes a few hours to
> transfer to the stand-alone network.  I shall do that tonight.  Is there a
> reason to expect this fixed in the latest snapshot when its been a problem
> for a long time (PR 207 is dated Tue Sep 04 09:18:01 PDT 2001) and I don't
> see any resolution information in the PRs.

A month or so back MichaelS KevinB and I gave the MIPS on IRIX stuff a 
``good talking too', n32, n64 and o32 now work a lot better.  It was all 
on 6.5 with the IRIX compiler, from memory 207 refers to 5.x.

Andrew

RE: PR 207, 662

[Smith, Jonathan C5]

Snapshot from 20020403, 5.2.0, 5.2.1.

I haven't taken a snapshot from the 5.3 branch as it takes a few hours to
transfer to the stand-alone network.  I shall do that tonight.  Is there a
reason to expect this fixed in the latest snapshot when its been a problem
for a long time (PR 207 is dated Tue Sep 04 09:18:01 PDT 2001) and I don't
see any resolution information in the PRs.

Jonathan Smith



-----Original Message-----
From: Andrew Cagney [mailto:ac131313@redhat.com]
Sent: Wednesday, October 23, 2002 2:46 PM
To: Smith, Jonathan C5
Cc: 'gdb@sources.redhat.com'
Subject: Re: PR 207, 662


Which GDB?  Have you tried a snap from the latest branch 
(ftp://sources.redhat.com/pub/gdb/snapshots/branch/)?

Andrew


> GDB has been unusable on SGI IRIX 6.5 for some time now.  The only useful
> information given by gdb is the following:
> 
> warning: Signal ? does not exist on this system.
> 
> But I know from attempts to debug that the hang is not immediately located
> around that.  Not knowing how gdb works, I could only get so far in
> attempting to debug this.  This hang occurs on any program I've attempted
to
> debug, not just the reported Python binary.
> 
> Is there anyone with any more information on this or some ability to solve
> this?
> 
> 
> Thank you,
> 
> Jonathan Smith
> --
> Jonathan C. Smith
> Voice:  (301) 342-6367
> Fax:     (301) 342-6381
> Atlantic Ranges & Facilities
> Computer Scientist
> Building 2109 
> 48150 Shaw Road, Suite S115
> Patuxent River, MD 20670-1907
> 
>  <<...OLE_Obj...>> 
> 
>