Multi-arch only GDB 6.0 ....

Multi-arch only GDB 6.0 ....

[Andrew Cagney]

So,

The GDB 5.0 version bump came about mainly because:

	o	5.0 requires an ISO C compiler

If someone wants to build GDB using a K&R compiler, they could use the 
older 4.x series or a ISO->K&R converter.  (It also was part of moving 
GDB to a public repository but that is secondary).

I think it is time to think about a GDB version bump.  This time the 
reason being (credit to Martin Hunt):

	o	6.0 only supports multi-arch targets

If someone wants to use GDB on an older more obscure target then they 
can grab an old 5.x release.

A tentative schedule (based on this 4 month release cycle) would be:

	jul-aug: 5.3 mark all as obsolete
	nov-dec: 6.0 have removed them (YA!)
	
Anyway, I'm going to mention all these going to be obsoleted targets in 
5.2's NEWS file.

Thoughts?
Andrew

Re: Multi-arch only GDB 6.0 ....

[Daniel Jacobowitz]

On Thu, Apr 18, 2002 at 08:08:45PM -0400, Andrew Cagney wrote:
> So,
> 
> The GDB 5.0 version bump came about mainly because:
> 
> 	o	5.0 requires an ISO C compiler
> 
> If someone wants to build GDB using a K&R compiler, they could use the 
> older 4.x series or a ISO->K&R converter.  (It also was part of moving 
> GDB to a public repository but that is secondary).
> 
> I think it is time to think about a GDB version bump.  This time the 
> reason being (credit to Martin Hunt):
> 
> 	o	6.0 only supports multi-arch targets
> 
> If someone wants to use GDB on an older more obscure target then they 
> can grab an old 5.x release.
> 
> A tentative schedule (based on this 4 month release cycle) would be:
> 
> 	jul-aug: 5.3 mark all as obsolete
> 	nov-dec: 6.0 have removed them (YA!)
> 	
> Anyway, I'm going to mention all these going to be obsoleted targets in 
> 5.2's NEWS file.
> 
> Thoughts?

It's a nice idea, but can we really ditch all the non-multi-arch targets?
I think you filed bugs about them all at one point; could you give an
exact list of the targets that would be going?

I'm stuck maintaining GDB for Debian right now.  In our current
release, that's 11 architectures.  I think that at the least Sparc,
Alpha, and m68k are non-multi-arch (and HPPA is non-contributed, but
there's nothing I can do about that...).  We may be able to talk David
into cleaning up Sparc, but at the other two are also still in active
use.  I think that we need to allow more time in order to get as many
targets as possible converted over - I'll do the two I mentioned myself
if no one else has time, but it'll take a while.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: Multi-arch only GDB 6.0 ....

[Andrew Cagney]

[-- Attachment #1: Type: text/plain, Size: 1187 bytes --]


> It's a nice idea, but can we really ditch all the non-multi-arch targets?
> I think you filed bugs about them all at one point; could you give an
> exact list of the targets that would be going?

Lets turn the question around.

Why should GDB 6.0 retain support non-multi-arch targets?  If someone 
needs to debug a non-multi-arch target they can download GDB 5.x.

For current status, see the file src/gdb/MAINTAINERS.  Hmm, the ARM 
entry is out-of-date!  The ones marked as ``OBSOLETE candidate'' are 
first up.

> I'm stuck maintaining GDB for Debian right now.  In our current
> release, that's 11 architectures.  I think that at the least Sparc,
> Alpha, and m68k are non-multi-arch (and HPPA is non-contributed, but
> there's nothing I can do about that...).  We may be able to talk David
> into cleaning up Sparc, but at the other two are also still in active
> use.  I think that we need to allow more time in order to get as many
> targets as possible converted over - I'll do the two I mentioned myself
> if no one else has time, but it'll take a while.

The SPARC is multi-arch, Alpha is WIP, m68k had an expression of 
interest, HP/UX is typically broken.

enjoy,
Andrew

[-- Attachment #2: MAINTAINERS --]
[-- Type: text/plain, Size: 2986 bytes --]


	a29k		OBSOLETE

	alpha		--target=alpha-dec-osf4.0a -Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	arc		--target=arc-elf ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	arm		--target=arm-elf -w
			Fernando Nasser		fnasser@redhat.com
			Scott Bambrough		scottb@netwinder.org
			Richard Earnshaw	rearnsha@arm.com
			Not multi-arch

	cris		--target=cris-elf -w
			Orjan Friberg		orjanf@axis.com

	d10v		--target=d10v-elf ,-Werror
			Maintenance only

	d30v		--target=d30v-elf ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	djgpp		--target=i586-pc-msdosdjgpp ,-Werror
			(See native and host)

	fr30		--target=fr30-elf -Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	h8300		--target=h8300hms -Werror
			Maintenance only
			Not multi-arch, work in progress

	h8500		--target=h8500hms -Werror
			Maintenance only
			Not multi-arch, work in progress

	i386		--target=i386-elf,i386-aout ,-Werror
			Mark Kettenis           kettenis@gnu.org

	i960		--target=i960-coff ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	ia64		--target=ia64-linux ,-Werror
			Kevin Buettner		kevinb@redhat.com

	m32r		--target=m32r-elf -Werror
			Michael Snyder		msnyder@redhat.com
			Not multi-arch

	m68hc11		--target=m68hc11-elf ,-Werror
			Stephane Carrez		Stephane.Carrez@worldnet.fr

	m68k		--target=m68k-elf ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	m88k		--target=m88k ,-Werror
			Known problem in 5.1
			Maintenance only
			OBSOLETE candidate, not multi-arch

	mcore		--target=mcore-elf,mcore-pe ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	mips		--target=mips-elf,mips64-elf ,-Werror
			Andrew Cagney		cagney@redhat.com

	mn10200		--target=mn10200-elf ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	mn10300		--target=mn10300-elf ,-Werror
			Maintenance only

	ns32k		--target=ns32k-netbsd ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	pa		(--target=hppa1.1-hp-proelf broken)
			Maintenance only
			OBSOLETE candidate, not multi-arch

	powerpc		--target=powerpc-eabi ,-Werror
			Kevin Buettner		kevinb@redhat.com

	rs6000		--target=rs6000-ibm-aix4.1 ,-Werror
			(see rs6000 native and ppc target)

	s390		--target=s390-linux ,-Werror
			(contact DJ Barrow	djbarrow@de.ibm.com)

	sh		--target=sh-hms,sh-elf -w
			Elena Zannoni		ezannoni@redhat.com

	sparc		--target=sparc-elf,sparc64-elf ,-Werror
			Maintenance only

	tic80		Deleted.

	v850		--target=v850-elf ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	vax		--target=vax-dec-vms5.5 ,-Werror
			Maintenance only
			OBSOLETE candidate, not multi-arch

	w65		Deleted.

	x86-64		(--target=x86_64-linux-gnu broken)
			Maintenance only

	xstormy16	--target=xstormy16-elf ,-Werror
			Corinna Vinschen	vinschen@redhat.com

	z8k		--target=z8k-coff ,-Werror
			Known problem in 5.1
			Maintenance only
			OBSOLETE candidate, not multi-arch

Re: Multi-arch only GDB 6.0 ....

[Daniel Jacobowitz]

On Thu, Apr 18, 2002 at 08:59:27PM -0400, Andrew Cagney wrote:
> 
> >It's a nice idea, but can we really ditch all the non-multi-arch targets?
> >I think you filed bugs about them all at one point; could you give an
> >exact list of the targets that would be going?
> 
> Lets turn the question around.
> 
> Why should GDB 6.0 retain support non-multi-arch targets?  If someone 
> needs to debug a non-multi-arch target they can download GDB 5.x.

Let me be clearer; I think there are some targets we need to allow time
to fix ourselves based on userbase, rather than on maintained-ness.  I
would wager that GDB on Alpha has a fair number of users, and it seems
to work relatively well.  If they prove unreasonably difficult to
multi-arch, well, that's a different story.

> For current status, see the file src/gdb/MAINTAINERS.  Hmm, the ARM 
> entry is out-of-date!  The ones marked as ``OBSOLETE candidate'' are 
> first up.
> 
> >I'm stuck maintaining GDB for Debian right now.  In our current
> >release, that's 11 architectures.  I think that at the least Sparc,
> >Alpha, and m68k are non-multi-arch (and HPPA is non-contributed, but
> >there's nothing I can do about that...).  We may be able to talk David
> >into cleaning up Sparc, but at the other two are also still in active
> >use.  I think that we need to allow more time in order to get as many
> >targets as possible converted over - I'll do the two I mentioned myself
> >if no one else has time, but it'll take a while.
> 
> The SPARC is multi-arch, Alpha is WIP, m68k had an expression of 
> interest, HP/UX is typically broken.

Most of SPARC is only multiarch if it snuck in with David's cleanups
recently.  A day-old tree:

tm-sp64.h:#define GDB_MULTI_ARCH GDB_MULTI_ARCH_PARTIAL
tm-sp64linux.h:#define GDB_MULTI_ARCH 0
tm-sun4sol2.h:#define GDB_MULTI_ARCH GDB_MULTI_ARCH_PARTIAL

Someone was working on HP/UX just a little while ago - Jeff Law, I
think.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: Multi-arch only GDB 6.0 ....

[Andrew Cagney]

> 
> Most of SPARC is only multiarch if it snuck in with David's cleanups
> recently.  A day-old tree:
> 
> tm-sp64.h:#define GDB_MULTI_ARCH GDB_MULTI_ARCH_PARTIAL
> tm-sp64linux.h:#define GDB_MULTI_ARCH 0

Hmm, I suspect that is a hangover from long ago.  David's patch needs to 
at least be multi-arch partial (I'd better check :-).

> tm-sun4sol2.h:#define GDB_MULTI_ARCH GDB_MULTI_ARCH_PARTIAL
> 
> Someone was working on HP/UX just a little while ago - Jeff Law, I
> think.

Jeff doesn't have the time to multi-arch HP/UX GDB.  Trust me on that :-)
Andrew

Re: Multi-arch only GDB 6.0 ....

[Eli Zaretskii]

> Date: Thu, 18 Apr 2002 20:59:27 -0400
> From: Andrew Cagney <ac131313@cygnus.com>
> 
> Why should GDB 6.0 retain support non-multi-arch targets?  If someone 
> needs to debug a non-multi-arch target they can download GDB 5.x.

That's not a practical solution, I'm afraid.  Old versions aren't
maintained, so they suffer bitrot very quickly as the world moves on.

For example, by the time GDB 6.0 is released, GDB 5.x might not even
compile with the then current development tools.  Are we going to tell
users of non-multi-arched targets to install old versions of
compilers, libraries, headers, etc.?

So in general, I'd advise against such a ``revolutionary'' approach.
In practice, it tends to be too harsh to users, IMHO.

Re: Multi-arch only GDB 6.0 ....

[Andreas Schwab]

Andrew Cagney <ac131313@cygnus.com> writes:

|> The SPARC is multi-arch, Alpha is WIP, m68k had an expression of interest,
|> HP/UX is typically broken.

I'd like to volunteer moving m68k to multi-arch.  Is there any docs on
what multi-arch is all about?

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE GmbH, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Multi-arch only GDB 6.0 ....

[Elena Zannoni]

Andreas Schwab writes:
 > Andrew Cagney <ac131313@cygnus.com> writes:
 > 
 > |> The SPARC is multi-arch, Alpha is WIP, m68k had an expression of interest,
 > |> HP/UX is typically broken.
 > 
 > I'd like to volunteer moving m68k to multi-arch.  Is there any docs on
 > what multi-arch is all about?
 > 

Yes, a few notes:

http://sources.redhat.com/gdb/papers/multi-arch/


Elena


 > Andreas.
 > 
 > -- 
 > Andreas Schwab, SuSE Labs, schwab@suse.de
 > SuSE GmbH, Deutschherrnstr. 15-19, D-90429 Nürnberg
 > Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
 > "And now for something completely different."

Re: Multi-arch only GDB 6.0 ....

[Andrew Cagney]

> Andreas Schwab writes:
>  > Andrew Cagney <ac131313@cygnus.com> writes:
>  > > |> The SPARC is multi-arch, Alpha is WIP, m68k had an expression of interest,
>  > |> HP/UX is typically broken.
>  > > I'd like to volunteer moving m68k to multi-arch.  Is there any docs on
>  > what multi-arch is all about?
>  > Yes, a few notes:
> 
> http://sources.redhat.com/gdb/papers/multi-arch/

The second (more up-to-date) source is the recent e-mail exchanges 
involving Richard E when they converted the ARM.

Andrew