Re: Orphans to adopt?

Re: Orphans to adopt?

[E. Weddington]

[Sorry for breaking any threading. Thunderbird stupidly doesn't allow me 
to reply to posts in digests.]

>
> Subject:
> Orphans to adopt?
> From:
> Stan Shebs <shebs@apple.com>
> Date:
> Tue, 17 May 2005 12:02:15 -0700
> To:
> gdb@sourceware.org
>
> To:
> gdb@sourceware.org
>
>
> After a month of distraction, I'm back on the Darwin GDB case,
> and for regression testing purposes it would be convenient to set
> up a random embedded target with a simulator. To make the fiddling
> do double duty, it occurs to me that I could adopt a target that
> is being neglected, although a cursory scan suggests that all
> adoptable targets are taken now. Does anybody have suggestions for
> a target in need of a little tenderness and cuddling? (hmm, sounds
> creepy that way :-) )
>
Well, I don't know if this would work for your purposes, but the AVR 
target could use some help. It's about as embedded as you can get. 
Forgive me if I don't know the correct terminology. There are 2 external 
simulators available for the AVR, both work with GDB and both are 
incomplete. What is desperately needed is a simulator that is 
"integrated" (?) with GDB (just like all the other, normal simulators) 
and is robust enough to run the GCC test suite. This would have an 
enormous impact on the ability of the AVR target in GCC to stay up to 
date and to test new improvements in that port that are trying to get 
online.

Eric
[ Please CC me in any replies.]

Orphans to adopt?

[Stan Shebs]

After a month of distraction, I'm back on the Darwin GDB case,
and for regression testing purposes it would be convenient to set
up a random embedded target with a simulator. To make the fiddling
do double duty, it occurs to me that I could adopt a target that
is being neglected, although a cursory scan suggests that all
adoptable targets are taken now. Does anybody have suggestions for
a target in need of a little tenderness and cuddling? (hmm, sounds
creepy that way :-) )

Stan

Re: Orphans to adopt?

[Daniel Jacobowitz]

On Tue, May 17, 2005 at 12:02:15PM -0700, Stan Shebs wrote:
> After a month of distraction, I'm back on the Darwin GDB case,
> and for regression testing purposes it would be convenient to set
> up a random embedded target with a simulator. To make the fiddling
> do double duty, it occurs to me that I could adopt a target that
> is being neglected, although a cursory scan suggests that all
> adoptable targets are taken now. Does anybody have suggestions for
> a target in need of a little tenderness and cuddling? (hmm, sounds
> creepy that way :-) )

Nothing jumps out at me.  If you just want one to test, might as well
use powerpc-eabisim.  The only one I'm working with lately is
arm-elf/arm-eabi, which is in adequate shape.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: Orphans to adopt?

[Mark Kettenis]

   Date: Tue, 17 May 2005 12:02:15 -0700
   From: Stan Shebs <shebs@apple.com>

   After a month of distraction, I'm back on the Darwin GDB case,
   and for regression testing purposes it would be convenient to set
   up a random embedded target with a simulator. To make the fiddling
   do double duty, it occurs to me that I could adopt a target that
   is being neglected, although a cursory scan suggests that all
   adoptable targets are taken now. Does anybody have suggestions for
   a target in need of a little tenderness and cuddling? (hmm, sounds
   creepy that way :-) )

Hmm, powerpc could really, really use some cleanup, but hardly
qualifies as a random embedded target in your case ;-).  Then there is
arm, which also uses some deprecated functionality that could use some
attention.

Mark

Re: Orphans to adopt?

[Manoj Iyer]

> Hmm, powerpc could really, really use some cleanup, but hardly

What type of clean up effort are we looking at for ppc. I am working on
ppc & linux and I can spend some time on this.

--
Manoj

Re: Orphans to adopt?

[Mark Kettenis]

   Date: Tue, 17 May 2005 15:51:47 -0500 (CDT)
   From: Manoj Iyer <manjo@austin.ibm.com>

   > Hmm, powerpc could really, really use some cleanup, but hardly

   What type of clean up effort are we looking at for ppc. I am working on
   ppc & linux and I can spend some time on this.

A couple of things:

* It still needs to be gdb_bytified.

* There are still a few deprecated things used by rs6000/powerpc.
  Look for things that have deprecated in their name.

* There is still a tm.h for most/all rs6000/powerpc targets.
  Basically config/powerpc/tm-ppc-eabi.h and config/rs6000/tm-rs6000.h
  need to be cleaned out.

* The code could use some reorganisation, especially the way the
  architecture vector is constructed.  Some of the OS/ABI-specefic
  stuff is handled in rs6000-tdep.c, some of it is dealt with in
  OS-specific files like ppc-linux-tdep.c and pccnbsd-tdep.c.

A lot of the mess seems to stem from the fact that we still try to
support AIX 4 & 5 on POWER as well as PowerPC.  But it seems that none
of the core gdb developers have access to hardware.

Mark

Re: Orphans to adopt?

[Andrew Cagney]

(Hmm, remind me, what was that system I used to identify the GCC 4 
gdb_byte (signed vs unsigned char) problems, that I've since been 
spending my time fixing?  Ah, yes ;-)

What would be immediatly helpful is for more people to expose GDB to GCC 
4 so that all the other problems above and beyond gdb_byte can be 
identified and fixed.

Re: Orphans to adopt?

[Mark Kettenis]

   Date: Thu, 19 May 2005 13:41:09 -0400
   From: Andrew Cagney <cagney@gnu.org>

   (Hmm, remind me, what was that system I used to identify the GCC 4 
   gdb_byte (signed vs unsigned char) problems, that I've since been 
   spending my time fixing?  Ah, yes ;-)

Yeah, but there are still some *structural* problems with
rs6000/powerpc; it's a bit of maze of twisted if's instead of clearly
seperated OSABI's.  A larche chunk of that poo is related to the old
POWER ABI; that could sure use some polishing by a loving soul.

   What would be immediatly helpful is for more people to expose GDB to GCC 
   4 so that all the other problems above and beyond gdb_byte can be 
   identified and fixed.

Yup, but GCC got a bit bloated and takes a long time too compile on my
poor old boxes.

Mark

Re: Orphans to adopt?

[Manoj Iyer]

>    What would be immediatly helpful is for more people to expose GDB to GCC
>    4 so that all the other problems above and beyond gdb_byte can be
>    identified and fixed.
>
> Yup, but GCC got a bit bloated and takes a long time too compile on my
> poor old boxes.
>

Yep, I am working on that in my spare time. We have internal daily builds
of GCC 4.0 which I am using to build and test GDB. I have one patch
outstanding that need to be approved for commit, its a patch for GDBserver
(there is a missing header file and wrong types). Can anyone look into
that please? SUB: " [RFC] GDBserver compiler warnings patch" gdbpatches@
pinged on June 6th 05.

What I can also do is, once a week/while post test results for
gdb+gcc4.0 to gdb-testers@. Is this useful?

Thanks
Manjo