Re: Which version of gdb supports gcc 3.0 ABI?

Re: Which version of gdb supports gcc 3.0 ABI?

[Christopher Faylor]

On Wed, Mar 14, 2001 at 01:25:00PM -0500, Phil Edwards wrote:
>I agree with Benjamin here: if they won't approve on a timely basis,
>and won't give you maintainer authority, then fork and do it the Right
>Way.

Daniel is the C++ maintainer for gdb.  The problem is that some of the
changes required to fix C++ handling touch on other code like the symbol
table parts of gdb.

I won't go into great details about why there were problems with patch
acceptance but suffice it to say that not all of the problems were due
to the fact that GDB patch approval is (arguably) slow.

So, before anyone draws conclusions on the GDB patch approval process,
please read the gdb and gdb-patches mailing list archives.

>That's great.  Well, it's not great that you're frustrated, but that
>you're rewriting it.  With years of stuff purged, someday I might be
>able to understand the debugger.  :-)

I have Cc'ed the gdb mailing list.  If you all have complaints about
gdb, it makes sense to talk about them there.

cgf

Re: Which version of gdb supports gcc 3.0 ABI?

[Daniel Berlin]

Christopher Faylor <cgf@redhat.com> writes:

> On Wed, Mar 14, 2001 at 01:25:00PM -0500, Phil Edwards wrote:
> >I agree with Benjamin here: if they won't approve on a timely basis,
> >and won't give you maintainer authority, then fork and do it the Right
> >Way.
> 
> Daniel is the C++ maintainer for gdb.  The problem is that some of the
> changes required to fix C++ handling touch on other code like the symbol
> table parts of gdb.

Having some good sense left, i'll just drop out of the discussion
at this point and get back to work. But there are plently  of patches
that  have nothing to do with C++, or even me.  It took a month to
approve Jason Merrill's simple AUTO_MANGLING change (I know it sounds
like a C++ fix, but it doesn't fall in my maintainership, so it's
not), for instance. There are plently  of examples.

> 
> I won't go into great details about why there were problems with patch
> acceptance but suffice it to say that not all of the problems were due
> to the fact that GDB patch approval is (arguably) slow.
And i'll leave this one alone. Suffice to say there are other problems
as well, but they are mostly minor compared to patch approval time.

> 
> So, before anyone draws conclusions on the GDB patch approval process,
> please read the gdb and gdb-patches mailing list archives.
Please, do.
Draw your own conclusions, don't rely on me, or anyone else. Try
submitting a patch for yourself (Hey, there's a cheap attempt to spur development).

> 
> >That's great.  Well, it's not great that you're frustrated, but that
> >you're rewriting it.  With years of stuff purged, someday I might be
> >able to understand the debugger.  :-)
> 
> I have Cc'ed the gdb mailing list.  If you all have complaints about
> gdb, it makes sense to talk about them there.
> 
Hey, Don't blame me,  Ben started it. :)
--Dan
> cgf

Re: Which version of gdb supports gcc 3.0 ABI?

[Eli Zaretskii]

> From: Daniel Berlin <dberlin@redhat.com>
> Date: 14 Mar 2001 22:35:09 -0500
> 
> Having some good sense left, i'll just drop out of the discussion
> at this point and get back to work. But there are plently  of patches
> that  have nothing to do with C++, or even me.  It took a month to
> approve Jason Merrill's simple AUTO_MANGLING change (I know it sounds
> like a C++ fix, but it doesn't fall in my maintainership, so it's
> not), for instance. There are plently  of examples.
> 
> > 
> > I won't go into great details about why there were problems with patch
> > acceptance but suffice it to say that not all of the problems were due
> > to the fact that GDB patch approval is (arguably) slow.
> And i'll leave this one alone. Suffice to say there are other problems
> as well, but they are mostly minor compared to patch approval time.

FWIW, I always had the GDB patches I posted as an RFA approved or
commented on in reasonable time (i.e. about a week).

In any case, a fork is not something people should consider easily,
just because someone's patch is not accepted quickly enough.

Re: Which version of gdb supports gcc 3.0 ABI?

[Steinar Bang]

>>>>> Eli Zaretskii <eliz@delorie.com>:

> FWIW, I always had the GDB patches I posted as an RFA approved or
> commented on in reasonable time (i.e. about a week).

> In any case, a fork is not something people should consider easily,
> just because someone's patch is not accepted quickly enough.

Maybe so.

But the lack of C++ support for gcc 3.0 is a problem.
And it will become a _huge_ problem when gcc 3.0 is released.

It must be adressed in some way, I think.  And a fork sounds like the
best alternative.

In a fork like this, one doesn't have to take into consideration
support for other languages in the common code.  Once the fork is
actually working for C++ in gcc 3.0, one can look into what changes
needs to be made in the common part of gdb, and argue for them, and
possibly work towards merging the new C++ support into the mainstream.

A fork doesn't _have_ to end up like GNU Emacs and XEmacs.  It could
be like with egcs and gcc (ie. first a fork and then a merge).

Re: Which version of gdb supports gcc 3.0 ABI?

[Eli Zaretskii]

> From: Steinar Bang <sb@metis.no>
> Date: 15 Mar 2001 19:50:03 +0100
> 
> But the lack of C++ support for gcc 3.0 is a problem.
> And it will become a _huge_ problem when gcc 3.0 is released.

As others told here, this support is being actively worked on, by
Daniel as well as by others.

> It must be adressed in some way, I think.  And a fork sounds like the
> best alternative.

IMHO, a fork is _never_ a good alternative, let alone the best one.

> A fork doesn't _have_ to end up like GNU Emacs and XEmacs.

How do you mean ``end up like GNU Emacs and XEmacs''?  That fork is
not over yet, so how will it end is anyone's guess.

Meanwhile, we do know what it does: it caused and continues to cause a
terrible waste of resources which are in short supply, whereby two
groups of talented and well-meaning individuals work against each
other instead of working together.  With large and complicated
packages such as Emacs and GDB, a fork is a dreadful blow to the
package development.

> It could be like with egcs and gcc (ie. first a fork and then a
> merge).

This can only happen if one of the branches is unmaintained.  That
possibility doesn't seem to be anywhere in sight wrt GDB.

A fork is an ultimate acknowledgment of a failure of people to
cooperate with each other.  Can anyone seriously say that this is what
happens in GDB maintenance?  If not, I suggest that we stop even
mentioning a fork, lest someone thinks we are being serious.

Re: Which version of gdb supports gcc 3.0 ABI?

[Andrew Cagney]

Steinar Bang wrote:
> 
> >>>>> Eli Zaretskii <eliz@delorie.com>:
> 
> > FWIW, I always had the GDB patches I posted as an RFA approved or
> > commented on in reasonable time (i.e. about a week).
> 
> > In any case, a fork is not something people should consider easily,
> > just because someone's patch is not accepted quickly enough.
> 
> Maybe so.
> 
> But the lack of C++ support for gcc 3.0 is a problem.
> And it will become a _huge_ problem when gcc 3.0 is released.
> 
> It must be adressed in some way, I think.  And a fork sounds like the
> best alternative.

Humor me.  What ``lack of C++ support for GCC 3.0''?

GDB 5.0 doesn't support the v3.0 ABI (Is this really surprising? - GDB
5.0 was released almost a year ago! :-).  GDB-current does provide ok
support for GCC's v3 ABI.

Michael Chastain has been pounding on GDB's C++ testsuite and hence
identifying real bugs both GDB and in v3.0 G++.  Jim Blandy has been
carefully revising Dan's proposed changes and seeing them into GDB/GCC. 
I assume Dan, as GDB's current C++ maintainer, is contributing to this
same effort.

In parallel to this, the GDB 5.1 release has been slowly picking up
speed.  It has already been decided that there will be a 5.1 and then a
likely 5.1.1.  The hope is that GDB 5.1 will come out before GCC 3.0 and
GDB 5.1.1 some time after. The intention is to have something sooner
rather than later.

	Have a good one!

		Andrew

Re: Which version of gdb supports gcc 3.0 ABI?

[Christopher Faylor]

On Thu, Mar 15, 2001 at 07:50:03PM +0100, Steinar Bang wrote:
>>>>>> Eli Zaretskii <eliz@delorie.com>:
>
>> FWIW, I always had the GDB patches I posted as an RFA approved or
>> commented on in reasonable time (i.e. about a week).
>
>> In any case, a fork is not something people should consider easily,
>> just because someone's patch is not accepted quickly enough.
>
>Maybe so.
>
>But the lack of C++ support for gcc 3.0 is a problem.
>And it will become a _huge_ problem when gcc 3.0 is released.
>
>It must be adressed in some way, I think.  And a fork sounds like the
>best alternative.

Again, I have to point out that DANIEL BERLIN IS THE C++ MAINTAINER for
gdb.  If people have issues with C++, they should be letting Daniel know
about them.

FYI, Daniel neglected to mention that we (Red Hat) have three people
working on the 3.0 problem (one of whom is him) and should have a
solution, soon.

I also again suggest that rather than jump up and down talking about
forks, go over to the gdb mailing lists and read about what is going
on.  Then you can draw opinions based on facts.

cgf

Re: Which version of gdb supports gcc 3.0 ABI?

[Eli Zaretskii]

> Date: Thu, 15 Mar 2001 15:21:27 -0500
> From: Christopher Faylor <cgf@redhat.com>
> 
> I also again suggest that rather than jump up and down talking about
> forks, go over to the gdb mailing lists and read about what is going
> on.  Then you can draw opinions based on facts.

Seconded.

Re: Which version of gdb supports gcc 3.0 ABI?

[Michael Elizabeth Chastain]

Hi Steinar,

> But the lack of C++ support for gcc 3.0 is a problem.
> And it will become a _huge_ problem when gcc 3.0 is released.

The current CVS version of gdb has some support for the gcc-3_0-branch
of gcc.  Several people are actively working on completing this.

Geoff Keating in the gcc group runs an automated gcc/gdb testing system.
You can see his results here:

  http://www.cygnus.com/~geoffk/gcc-regression/

I post a weekly build-and-test report here:

  http://sources.redhat.com/ml/gdb-testers

If you have specific bugs, the gdb Gnats database is open for bug reports.
We need to know all the usual stuff:

  your platform
  the version of gcc you are using
  the version of gdb you are using
  your test program
  the behavior you are seeing

Basically, if you file bugs with "gcc snapshot 2001-XX-XX" and "gdb CVS
version 2001-YY-YY", then we are interested.  I've filed several such
bugs and people are working on them.

Michael Elizabeth Chastain
<chastain@redhat.com>
"love without fear"