Re: my 6.0 todo list

Re: my 6.0 todo list

[Michael Elizabeth Chastain]

elena> I didn't have a chance to continue on this, sorry.

Mmmm okay ... anyone want to pick up on this?  You need a red hat linux
9 box, about 10 gigabytes of disk space, and a 36 hour script run (on
an 800 MHz Celeron with 128 MB memory, will be less if you have a
faster computer).

elena> I'd like to add this to the list of gcc's we test. I hope I'll get a
elena> chance to do this soon. There is a public branch that is called
elena> tree-ssa-20020619-branch.

I'll throw that on my TODO list.  In a week or so the report will pop
out.

gcc just left stage 1, which is the stage where they can import branches
into HEAD.  They will be in stage 2 for two months (development on HEAD)
followed by stage 3 for two months (bug fixes on HEAD) followed by
cutting gcc-3_4-branch.  Then they go back into stage 1 and SSA can be
imported into HEAD while they prepare gcc 3.4.

That's their road map on paper, anyways.

Michael C

Re: my 6.0 todo list

[Michael Elizabeth Chastain]

Elena Z writes:
> I'd like to add this to the list of gcc's we test. I hope I'll get a
> chance to do this soon. There is a public branch that is called
> tree-ssa-20020619-branch.

Okay, I tossed tree-ssa-20020619-branch into my scripts for the current
spin.

I have 128 megabytes of memory, and I've got a 175 megabyte cc1plus
running.  It's swapping the crap out of my poor IDE disk!

The command is: xgcc libjava/interpreter.cc.  Not that it matters.  It's
been running for 90 minutes.  I expect that it's spending most of its
time banking on the swap partition.

So, anyone who wants to play with the ssa branch, get a machine with 256
megabytes or more of memory (or maybe a 192 megabyte machine and don't
run X or anything else on it).

tree-ssa is going in the "do not care until after gdb 6.0 is release"
category for me.

Michael C

Re: my 6.0 todo list

[Andrew Cagney]

>> 
>> Just fyi,
>> 
>> In addition to the usual stuff I need to clear the following backlog:
>> 
>> - m32r revival
>> - annotate 3 commit
>> - mi 2->3 switch (needed?)
> 
> 
> Are the changes non-backward compatible ?
> Did mi2 ever went out in any official release ?
> 
> 5.2.1 & 5.3 was -mi 1
> 
> Sorry I did not follows the in/out of MI.  Feel free to redirect
> to the appropriate README/NEWS.

I need to audit exactly what, in MI land, changed.  If there is 
something lumpy then the following should happen:

cp mi-*.exp mi2-*.exp to `freeze' mi2.
increment mi2->mi3 in the mainline.

Andrew

Re: my 6.0 todo list

[Andrew Cagney]

> 
> Are the changes non-backward compatible ?
> Did mi2 ever went out in any official release ?
> 
> 5.2.1 & 5.3 was -mi 1
> 
> Sorry I did not follows the in/out of MI.  Feel free to redirect
> to the appropriate README/NEWS.
> 
> I need to audit exactly what, in MI land, changed.  If there is something lumpy then the following should happen:
> 
> cp mi-*.exp mi2-*.exp to `freeze' mi2.
> increment mi2->mi3 in the mainline.
> 
> Andrew 

Ok, `I remember'.

Post this release, the varobj interfaces will need to be changed so that 
it uses frame-id's, and not simple stack addresses, to identify specific 
frames.  It will fix some nasty edge cases but at the same time (likely) 
break compatibility.  The old way was simply wrong :-/

Andrew

Re: my 6.0 todo list

[Michael Elizabeth Chastain]

Hi Michal,

> Pretty good. Now I have only 34 unexpected failures here.

Would you say that it's in a releasable state then?

You can send me a gdb.log if you want me to scrutinize it.

(Also wondering if this is any use to Andrew, or if I'm just stepping on
his toes).

Michael C

Re: my 6.0 todo list

[Michal Ludvig]

Michael Elizabeth Chastain told me that:
> Hi Michal,
> 
>>Pretty good. Now I have only 34 unexpected failures here.
> 
> Would you say that it's in a releasable state then?

Yes it is. Many thanks to MarkK who integrated i386 and amd64 targets!

> You can send me a gdb.log if you want me to scrutinize it.

PLease take the recent logs from: http://tmp.logix.cz/amd/gdb/log/

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* (+420) 296.545.373 * http://www.suse.cz

Re: my 6.0 todo list

[Michael Elizabeth Chastain]

On my list:

  gdb backtrace bugs
    http://sources.redhat.com/gdb/bugs/1250
    http://sources.redhat.com/gdb/bugs/1253
    http://sources.redhat.com/gdb/bugs/1255

  dejagnu build issue
    http://sources.redhat.com/gdb/bugs/708

  red hat linux 9
    elena z how is it going?

Off the top of my head, my test bed isn't producing any regressions
besides the backtrace bugs.  But I won't swear that it is clean until
I get out the microscope though.

From my last "5.3 versus HEAD" report,

  MI issues
    how is MI these days?

  x86-64 regressions
    how is it these days?

  multi-register variables
    fixed!

  java regression
    fixed!

  gcc HEAD -gstabs+
    The nastiest pr, gcc/10055, got fixed in gcc.

    There are other regressions with gcc HEAD with both dwarf-2 and
    stabs+ but this does not stop a gdb release.

I will put out a "5.3 versus gdb_6_0-branch" report next week.
I've been hoping for some of the backtrace bugs to get fixed first.

Michael C

Re: my 6.0 todo list

[Michal Ludvig]

Michael Elizabeth Chastain told me that:
>   x86-64 regressions
>     how is it these days?

Pretty good. Now I have only 34 unexpected failures here.

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* (+420) 296.545.373 * http://www.suse.cz

Re: my 6.0 todo list

[Elena Zannoni]

Michael Elizabeth Chastain writes:
 > On my list:
 > 
 >   gdb backtrace bugs
 >     http://sources.redhat.com/gdb/bugs/1250
 >     http://sources.redhat.com/gdb/bugs/1253
 >     http://sources.redhat.com/gdb/bugs/1255
 > 
 >   dejagnu build issue
 >     http://sources.redhat.com/gdb/bugs/708
 > 
 >   red hat linux 9
 >     elena z how is it going?
 > 

I didn't have a chance to continue on this, sorry.

It occured to me that there is a potential train about to hit us,
which is the tree-ssa gcc work. Apparently it is going to produce
really really heavily optimized code, and I am worried abuot how well
gdb will cope for this.

I'd like to add this to the list of gcc's we test. I hope I'll get a
chance to do this soon. There is a public branch that is called
tree-ssa-20020619-branch.

elena


 > Off the top of my head, my test bed isn't producing any regressions
 > besides the backtrace bugs.  But I won't swear that it is clean until
 > I get out the microscope though.
 > 
 > >From my last "5.3 versus HEAD" report,
 > 
 >   MI issues
 >     how is MI these days?
 > 
 >   x86-64 regressions
 >     how is it these days?
 > 
 >   multi-register variables
 >     fixed!
 > 
 >   java regression
 >     fixed!
 > 
 >   gcc HEAD -gstabs+
 >     The nastiest pr, gcc/10055, got fixed in gcc.
 > 
 >     There are other regressions with gcc HEAD with both dwarf-2 and
 >     stabs+ but this does not stop a gdb release.
 > 
 > I will put out a "5.3 versus gdb_6_0-branch" report next week.
 > I've been hoping for some of the backtrace bugs to get fixed first.
 > 
 > Michael C

Re: my 6.0 todo list

[Daniel Jacobowitz]

On Thu, Jul 17, 2003 at 11:54:05AM -0400, Elena Zannoni wrote:
> Michael Elizabeth Chastain writes:
>  > On my list:
>  > 
>  >   gdb backtrace bugs
>  >     http://sources.redhat.com/gdb/bugs/1250
>  >     http://sources.redhat.com/gdb/bugs/1253
>  >     http://sources.redhat.com/gdb/bugs/1255
>  > 
>  >   dejagnu build issue
>  >     http://sources.redhat.com/gdb/bugs/708
>  > 
>  >   red hat linux 9
>  >     elena z how is it going?
>  > 
> 
> I didn't have a chance to continue on this, sorry.
> 
> It occured to me that there is a potential train about to hit us,
> which is the tree-ssa gcc work. Apparently it is going to produce
> really really heavily optimized code, and I am worried abuot how well
> gdb will cope for this.
> 
> I'd like to add this to the list of gcc's we test. I hope I'll get a
> chance to do this soon. There is a public branch that is called
> tree-ssa-20020619-branch.

It isn't till 3.5; by then I hope to have... something... anything...
better able to cope than what we have now.  The way my luck's been
going we'll have to wait and see.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

my 6.0 todo list

[Andrew Cagney]

Just fyi,

In addition to the usual stuff I need to clear the following backlog:

- m32r revival
- annotate 3 commit
- mi 2->3 switch (needed?)

You'll note that none of these are highly technical.

Andrew

Re: my 6.0 todo list

[Alain Magloire]

> 
> Just fyi,
> 
> In addition to the usual stuff I need to clear the following backlog:
> 
> - m32r revival
> - annotate 3 commit
> - mi 2->3 switch (needed?)

Are the changes non-backward compatible ?
Did mi2 ever went out in any official release ?

5.2.1 & 5.3 was -mi 1

Sorry I did not follows the in/out of MI.  Feel free to redirect
to the appropriate README/NEWS.