Re: branch comparison tables

Re: branch comparison tables

[Michael Elizabeth Chastain]

> By the way, FYI: At this moment, the FAILs (only, not looking at the
> kfails at the moment) for GCC 3.4 and GCC HEAD are in several
> categories:

Well, that's what "compare by gcc" is for.

In this set of tables:

  http://www.shout.net/~mec/sunday/2004-01-19-branch

Here are the files that have a regression from gcc 3.3.2 to gcc
gcc-3_4-branch or a regression from gcc 3.3.2 to gcc HEAD,
with either gdb 6.0 or gdb HEAD:

  gdb.base/break.exp
  gdb.base/scope.exp
  gdb.base/sepdebug.exp
  gdb.base/store.exp
  gdb.cp/anon-union.exp
  gdb.cp/classes.exp
  gdb.cp/local.exp
  gdb.cp/namespace.exp
  gdb.java/jmisc1.exp
  gdb.java/jmisc2.exp
  gdb.mi/mi-until.exp
  gdb.mi/mi-var-block.exp
  gdb.mi/mi1-until.exp
  gdb.mi/mi1-var-block.exp
  gdb.mi/mi2-until.exp
  gdb.mi/mi2-var-block.exp
  gdb.threads/tls.exp

> I believe that no more of them are caused by GCC.

PR gcc/12267 is still causing stabs problems.
gdb.cp/anon-union.exp is blowing up, I haven't analyzed it yet.

> The multi_line_while_statement failures in your tables were, but I
> checked in a fix this morning.

That will show up in my next spin, 36-48 hours from now.

Michael C

Re: branch comparison tables

[Daniel Jacobowitz]

On Thu, Jan 22, 2004 at 10:38:54PM -0500, Michael Chastain wrote:
> > By the way, FYI: At this moment, the FAILs (only, not looking at the
> > kfails at the moment) for GCC 3.4 and GCC HEAD are in several
> > categories:
> 
> Well, that's what "compare by gcc" is for.
> 
> In this set of tables:
> 
>   http://www.shout.net/~mec/sunday/2004-01-19-branch
> 
> Here are the files that have a regression from gcc 3.3.2 to gcc
> gcc-3_4-branch or a regression from gcc 3.3.2 to gcc HEAD,
> with either gdb 6.0 or gdb HEAD:
> 
>   gdb.base/break.exp
>   gdb.base/scope.exp
>   gdb.base/sepdebug.exp
>   gdb.base/store.exp
>   gdb.cp/anon-union.exp
>   gdb.cp/classes.exp
>   gdb.cp/local.exp
>   gdb.cp/namespace.exp
>   gdb.java/jmisc1.exp
>   gdb.java/jmisc2.exp
>   gdb.mi/mi-until.exp
>   gdb.mi/mi-var-block.exp
>   gdb.mi/mi1-until.exp
>   gdb.mi/mi1-var-block.exp
>   gdb.mi/mi2-until.exp
>   gdb.mi/mi2-var-block.exp
>   gdb.threads/tls.exp
> 
> > I believe that no more of them are caused by GCC.
> 
> PR gcc/12267 is still causing stabs problems.

Sorry, I forgot to mention above - I only did DWARF2.

> gdb.cp/anon-union.exp is blowing up, I haven't analyzed it yet.

It's a legitimate change in the debug output that GDB is not prepared
for.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

branch comparison tables

[Michael Elizabeth Chastain]

Here are some new tables:

  http://www.shout.net/~mec/sunday/2004-01-19-branch

This is all four branches (6.0, HEAD, carlton_dictionary-branch,
drow-cplus-branch) run with the test suite from gdb HEAD.
So the compare-by-gdb tables are actually useful here.

Looking at compare-by-gdb, at the sub-tables with "as=2.14, ld=2.14".
There is not a lot of variation based on the version of binutils.
So sub-tables #2 #3 #6 #7 #10 #11 #16 #17 #22 #23 #28 #29 ought to
cover all the interesting information.

I would like to do more of this.  Conceptually, I've got it worked out
that the version of the gdb test suite and the version of gdb are
independent attributes of a test run.  I just have to figure out how to
present it usefully, and then write some more Perl.

Michael C

Re: branch comparison tables

[Daniel Jacobowitz]

On Thu, Jan 22, 2004 at 06:07:58PM -0500, Michael Chastain wrote:
> Here are some new tables:
> 
>   http://www.shout.net/~mec/sunday/2004-01-19-branch
> 
> This is all four branches (6.0, HEAD, carlton_dictionary-branch,
> drow-cplus-branch) run with the test suite from gdb HEAD.
> So the compare-by-gdb tables are actually useful here.
> 
> Looking at compare-by-gdb, at the sub-tables with "as=2.14, ld=2.14".
> There is not a lot of variation based on the version of binutils.
> So sub-tables #2 #3 #6 #7 #10 #11 #16 #17 #22 #23 #28 #29 ought to
> cover all the interesting information.
> 
> I would like to do more of this.  Conceptually, I've got it worked out
> that the version of the gdb test suite and the version of gdb are
> independent attributes of a test run.  I just have to figure out how to
> present it usefully, and then write some more Perl.

Neat.

By the way, FYI: At this moment, the FAILs (only, not looking at the
kfails at the moment) for GCC 3.4 and GCC HEAD are in several
categories:
  - gdb bugs
  - testsuite bugs
  - java tests, blah
  - namespace issue that David is looking at

I believe that no more of them are caused by GCC.  The
multi_line_while_statement failures in your tables were, but I checked
in a fix this morning.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer