GDB 6.2.1?

GDB 6.2.1?

[Andrew Cagney]

Hello,

What's left for this re-spin?
- ``absolute source path'' patch
- i386 patch
- ?

Andrew

Re: GDB 6.2.1?

[Michael Chastain]

Andrew Cagney <cagney@gnu.org> wrote:
> What's left for this re-spin?

I haven't seen anything gruesome from my test bed lately, except for all
the gcc-HEAD stuff.

(In general I don't feel a need to backport test suite changes anyways,
because I just run gdb gdb_6_2-branch with the suite from HEAD.  It
would be nice if I made it easier for other people to do this too.)

Michael C

Re: GDB 6.2.1?

[Joel Brobecker]

> What's left for this re-spin?
> - ``absolute source path'' patch
> - i386 patch
> - ?

2 mips-irix patches. You said in one of your earlier message that I
should feel free to pull in the branch any patch I need for mips-irix.
I understood that these patches needed to be checked in mainline first
before they could be integrated in the branch. That's why there not
in the branch yet.

The first patch was under heavy discussion (about setting the right
floatformat). I am not where we're headed for this one as I don't have
the time to work on adding support for IRIX 128bit floats in GDB.

The other one needs to be reworked because Andrew wanted to get some
reorg done before we looked into this. It's on my plate.

What should we do? Push a bit on the resolution of the above 2 patches?
Or simply pull them in the branch as is, without having checked
something in head first (this is unorthodox)?

Cheers,
-- 
Joel

Re: GDB 6.2.1?

[Andrew Cagney]

>>What's left for this re-spin?
>>> - ``absolute source path'' patch
>>> - i386 patch
>>> - ?

The MIPS crasher bug is in though.  That's the critical one.

> 2 mips-irix patches. You said in one of your earlier message that I
> should feel free to pull in the branch any patch I need for mips-irix.
> I understood that these patches needed to be checked in mainline first
> before they could be integrated in the branch. That's why there not
> in the branch yet.
> 
> The first patch was under heavy discussion (about setting the right
> floatformat). I am not where we're headed for this one as I don't have
> the time to work on adding support for IRIX 128bit floats in GDB.

I think it ended with a compromise of a 128-bit IRIX MIPS floatformat.

I'd also like to know the original details on the crash.

> The other one needs to be reworked because Andrew wanted to get some
> reorg done before we looked into this. It's on my plate.

I don't even remember what that one is.  Ah, yes, the backtrace fix, 
leave that for 6.3.

> What should we do? Push a bit on the resolution of the above 2 patches?
> Or simply pull them in the branch as is, without having checked
> something in head first (this is unorthodox)?

(We've all learnt the hard way to not do this).

So lets get just cover the float bug.

Andrew

Re: GDB 6.2.1?

[Joel Brobecker]

> >>What's left for this re-spin?
> >>>- ``absolute source path'' patch
> >>>- i386 patch
> >>>- ?
> 
> The MIPS crasher bug is in though.  That's the critical one.

Actually, no, the crasher bug is still not fixed (it's the one
dealing with the size of long doubles). The patch that we checked in
was fixing "break main".

> >The first patch was under heavy discussion (about setting the right
> >floatformat). I am not where we're headed for this one as I don't have
> >the time to work on adding support for IRIX 128bit floats in GDB.
> 
> I think it ended with a compromise of a 128-bit IRIX MIPS floatformat.
> I'd also like to know the original details on the crash.

OK, so I'll try to work on that asap and send a patch. I'll resend the
details of the crash.

> >The other one needs to be reworked because Andrew wanted to get some
> >reorg done before we looked into this. It's on my plate.
> 
> I don't even remember what that one is.  Ah, yes, the backtrace fix, 
> leave that for 6.3.

It fixes the backtrace problem, but the problem it fixes also influences
next/step.

> >What should we do? Push a bit on the resolution of the above 2 patches?
> >Or simply pull them in the branch as is, without having checked
> >something in head first (this is unorthodox)?
> 
> (We've all learnt the hard way to not do this).

I agree. It's just I wasn't sure about what you meant in your message.

-- 
Joel

Re: GDB 6.2.1?

[Andrew Cagney]

>>>>> >>What's left for this re-spin?
>>>>
>>>>>> >>>- ``absolute source path'' patch
>>>>>> >>>- i386 patch
>>>>>> >>>- ?
>>
>>> 
>>> The MIPS crasher bug is in though.  That's the critical one.
> 
> 
> Actually, no, the crasher bug is still not fixed (it's the one
> dealing with the size of long doubles). The patch that we checked in
> was fixing "break main".

``break main'' was, for me the crasher bug :-)  The criteria explicitly 
mention ``does it fix "break main; run"'' which it did :-)

As for the FP crash, thanks to your bug analysis there's something to 
stop it in the branch and a mainline fix is well in hand.

>>> I don't even remember what that one is.  Ah, yes, the backtrace fix, 
>>> leave that for 6.3.
> 
> 
> It fixes the backtrace problem, but the problem it fixes also influences
> next/step.

I still think we're better off shooting for a decent clean-up by 6.3 (in 
a few months).

I'll commit the ``absolute source path'' patch then 6.2.1 is good to go 
(well depending on how Eli and Mark resolve the i386 change).

Andrew

Re: GDB 6.2.1?

[Andrew Cagney]

> I'll commit the ``absolute source path'' patch then 6.2.1 is good to go (well depending on how Eli and Mark resolve the i386 change). 

I'm going to back out of my decision here :-(

I've looked over the patch and am really reluctant to back-port it 
without a testcase illustrating what it fixed.  While yes a bug, its not 
like the absolute shocker MIPS crash in 6.2.  It's also long standing. 
I think for this we should instead accelerate 6.3 a little and pull it 
into the Oct/Nov timeframe.

Andrew

Re: GDB 6.2.1?

[Eli Zaretskii]

> Date: Sat, 07 Aug 2004 14:31:30 -0400
> From: Andrew Cagney <cagney@gnu.org>
> 
> > I'll commit the ``absolute source path'' patch then 6.2.1 is good to go (well depending on how Eli and Mark resolve the i386 change). 
> 
> I'm going to back out of my decision here :-(
> 
> I've looked over the patch and am really reluctant to back-port it 
> without a testcase illustrating what it fixed.

If you are talking about the ``absolute source path'' patch, then I
tend to agree that 6.2.1 could do without it.

Re: GDB 6.2.1?

[Baurjan Ismagulov]

Hello,

On Sat, Aug 07, 2004 at 02:31:30PM -0400, Andrew Cagney wrote:
> >I'll commit the ``absolute source path'' patch then 6.2.1 is good to go 
> >(well depending on how Eli and Mark resolve the i386 change). 
> 
> I'm going to back out of my decision here :-(
> 
> I've looked over the patch and am really reluctant to back-port it 
> without a testcase illustrating what it fixed.

I've tried the test as given in
http://sources.redhat.com/ml/gdb-patches/2004-04/msg00463.html , it
doesn't work with 6.2 and does work with 6.2 + patch.

As for testsuite entry, it is on my TODO list for a long time. I'll try
to do it these weeks.


> While yes a bug, its not 
> like the absolute shocker MIPS crash in 6.2.  It's also long standing. 
> I think for this we should instead accelerate 6.3 a little and pull it 
> into the Oct/Nov timeframe.

No problem with me.


With kind regards,
Baurjan.

Re: GDB 6.2.1?

[Andrew Cagney]

> Here's how things are panning out:
> 
> week 0.0: @6.2:
> For current mainstream systems, our best releaes ever!
> Late breaking discovery that MIPS is broken, time to resolution unknown (guess 2 weeks) but workaround in hand.
> 
> week 1.5: 6.2.1:
> Critical MIPS problem fixed.
> Fix for long-standing i386 bug known (needs 2 weeks testing)

I'm going to drop 6.2.1.  I think we've now missed the oportunity.  It 
was ment to go out last weekend and instead it is now within a stones 
throw of ...

> week 4.0: 6.2.2:
> Long standing bug on old i386 systems fixed. 

... I'll put this down for around 30th of august (I'm also heading into 
an extreemly hectic week).

Andrew