Register setting problem

Register setting problem

[Ron McCall]

Hi,

I am using GDB from CVS on August 12th, targeted to powerpc-eabi.  No
matter which general purpose register I try to set, r0 gets set instead.
For example, both "set $r3=15" and "set var $r3=15" result in r0==15 and
r3 unchanged.  In case it matters, I am using target remote.  With
remote debugging turned on, I can see 'P' packets go out in these cases
but always with register number 0.  If I disable the 'P' packet, a 'G'
packet goes out instead but the modified value is still the first 8 hex
digits (register r0).

Has anyone else seen this?

Ron McCall

Re: Register setting problem

[Daniel Jacobowitz]

On Tue, Sep 17, 2002 at 05:11:41PM -0400, Ron McCall wrote:
> Hi,
> 
> I am using GDB from CVS on August 12th, targeted to powerpc-eabi.  No
> matter which general purpose register I try to set, r0 gets set instead.
> For example, both "set $r3=15" and "set var $r3=15" result in r0==15 and
> r3 unchanged.  In case it matters, I am using target remote.  With
> remote debugging turned on, I can see 'P' packets go out in these cases
> but always with register number 0.  If I disable the 'P' packet, a 'G'
> packet goes out instead but the modified value is still the first 8 hex
> digits (register r0).
> 
> Has anyone else seen this?

Yes.  I believe you need a newer snapshot; Andrew fixed it soon after.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: Register setting problem

[Ron McCall]

On Tue, Sep 17, 2002 at 05:15:44PM -0400, Daniel Jacobowitz wrote:
> Yes.  I believe you need a newer snapshot; Andrew fixed it soon after.

OK, I will build a new version and give it a try!  Thanks!

Ron McCall

Re: Register setting problem

[Andrew Cagney]

> On Tue, Sep 17, 2002 at 05:15:44PM -0400, Daniel Jacobowitz wrote:
> 
>> Yes.  I believe you need a newer snapshot; Andrew fixed it soon after.

(yes, I think I fixed it :-)

> OK, I will build a new version and give it a try!  Thanks!

Suggest grabbing a snapshot from the latest (5.3) branch: 
ftp://sources.redhat.com/pub/gdb/snapshots/branch/.  That way we know it 
is fixed in the next GDB release.

Andrew

Re: Register setting problem

[Ron McCall]

On Tue, Sep 17, 2002 at 06:03:14PM -0400, Andrew Cagney wrote:
> Suggest grabbing a snapshot from the latest (5.3) branch: 
> ftp://sources.redhat.com/pub/gdb/snapshots/branch/.  That way we know it 
> is fixed in the next GDB release.
> 
> Andrew

Good timing!  I was just pulling the latest from CVS and wasn't sure if
I should have used the 5.3 branch instead or not.  I will go ahead and
grab the latest 5.3 snapshot and test it out.  Thanks!

Ron McCall

Re: Register setting problem

[Ron McCall]

On Tue, Sep 17, 2002 at 06:03:14PM -0400, Andrew Cagney wrote:
> Suggest grabbing a snapshot from the latest (5.3) branch: 
> ftp://sources.redhat.com/pub/gdb/snapshots/branch/.  That way we know
> it is fixed in the next GDB release.

OK, I grabbed gdb+dejagnu.tar.bz2 which appears to be symlinked to
gdb+dejagnu-5.2.90_20020917.tar.bz2.  I am getting a configure error:

checking whether make sets ${MAKE}... (cached) yes
configure: error: source directory already configured; run make
distclean there first
configure: error:
.../../../../gdb+dejagnu-5.2.90_20020917/dejagnu/example/calc/configure
failed for example/calc
Configure in /export/home0/software/build/build-gdb/dejagnu failed,
exiting.

The config.log file in the calc subdirectory looks clean (though there
is a confdefs.h file left over containing only a single newline
character):

This file contains any messages produced by compilers while
running configure, to aid debugging if configure makes a mistake.

configure:562: checking for a BSD compatible install
configure:615: checking whether build environment is sane
configure:672: checking whether make sets ${MAKE}

I then tried the suggested action of running make distclean in the GDB
source directory but that results in:

make: *** No rule to make target `distclean'.  Stop.

I get the same error in the dejagnu subdirectory but I get a different
error in the dejagnu/example subdirectory:

Makefile:317: warning: overriding commands for target `check-recursive'
Makefile:111: warning: ignoring old commands for target
`check-recursive'
make: *** No rule to make target `../config.status', needed by
`Makefile'.  Stop.

In the dejagnu/example/calc subdirectory, I get the same error minus
the two warnings.

Does anyone know how to get around this problem?  Should I try
yesterday's snapshot instead?  I am building GDB in a separate directory
outside of the source tree as I usually do and the source tree is clean
from the tarball.

Thanks for the help!

Ron McCall

Re: Register setting problem

[Andrew Cagney]

> On Tue, Sep 17, 2002 at 06:03:14PM -0400, Andrew Cagney wrote:
> 
>> Suggest grabbing a snapshot from the latest (5.3) branch: 
>> ftp://sources.redhat.com/pub/gdb/snapshots/branch/.  That way we know
>> it is fixed in the next GDB release.
> 
> 
> OK, I grabbed gdb+dejagnu.tar.bz2 which appears to be symlinked to
> gdb+dejagnu-5.2.90_20020917.tar.bz2.  I am getting a configure error:
> 
> checking whether make sets ${MAKE}... (cached) yes
> configure: error: source directory already configured; run make
> distclean there first
> configure: error:
> .../../../../gdb+dejagnu-5.2.90_20020917/dejagnu/example/calc/configure
> failed for example/calc
> Configure in /export/home0/software/build/build-gdb/dejagnu failed,
> exiting.

I think the snapshot (and hence the release :-() process is leaving a 
stray file behind. (I also recall someone mentioning this before :-( ).

Look for config.cache, config.log, Makefile, or similar.

The work around is to delete the dejagnu directory from the unpacked 
source tree.

Andrew


> The config.log file in the calc subdirectory looks clean (though there
> is a confdefs.h file left over containing only a single newline
> character):
> 
> This file contains any messages produced by compilers while
> running configure, to aid debugging if configure makes a mistake.
> 
> configure:562: checking for a BSD compatible install
> configure:615: checking whether build environment is sane
> configure:672: checking whether make sets ${MAKE}
> 
> I then tried the suggested action of running make distclean in the GDB
> source directory but that results in:
> 
> make: *** No rule to make target `distclean'.  Stop.
> 
> I get the same error in the dejagnu subdirectory but I get a different
> error in the dejagnu/example subdirectory:
> 
> Makefile:317: warning: overriding commands for target `check-recursive'
> Makefile:111: warning: ignoring old commands for target
> `check-recursive'
> make: *** No rule to make target `../config.status', needed by
> `Makefile'.  Stop.
> 
> In the dejagnu/example/calc subdirectory, I get the same error minus
> the two warnings.
> 
> Does anyone know how to get around this problem?  Should I try
> yesterday's snapshot instead?  I am building GDB in a separate directory
> outside of the source tree as I usually do and the source tree is clean
> from the tarball.
> 
> Thanks for the help!
> 
> Ron McCall
> 

Re: Register setting problem

[Ron McCall]

On Tue, Sep 17, 2002 at 07:55:57PM -0400, Andrew Cagney wrote:
> I think the snapshot (and hence the release :-() process is leaving a 
> stray file behind. (I also recall someone mentioning this before :-( ).
> 
> Look for config.cache, config.log, Makefile, or similar.
> 
> The work around is to delete the dejagnu directory from the unpacked 
> source tree.

Getting rid of Makefile, config.log and config.status from the entire
source tree did the trick.  I am now building.  Thanks for the help!

Ron McCall

Re: Register setting problem

[Ron McCall]

On Tue, Sep 17, 2002 at 06:03:14PM -0400, Andrew Cagney wrote:
> Suggest grabbing a snapshot from the latest (5.3) branch:
> ftp://sources.redhat.com/pub/gdb/snapshots/branch/.  That way we know
> it is fixed in the next GDB release.

Just wanted to report back that setting registers in 5.2.90_20020917
works.  This was for the powerpc-eabi target.

Thanks for the help!

Ron McCall