Re: [Gdb] removing src/dejagnu, src/expect

Re: [Gdb] removing src/dejagnu, src/expect

[Michael Elizabeth Chastain]

Jim Ingham writes:
> I just reran the gdb testsuite - on the Apple gdb but this shouldn't 
> make much difference - with the Expect 5.38 from the SourceForge 
> sources, and with Tcl 8.4.4, also from SourceForge, and I didn't see 
> any new failures compared with the expect & tcl from src/.

I've done a similar comparison with the gdb testsuite on native
i686-pc-linux-gnu, (red hat 8.0, LANG=en_US.UTF-8), and the results
came out the same.  I've been using the released versions of
tcl+expect+dejagnu for 2.5 years now.

That's just gdb experience though.

I'm okay with shooting first and fixing problems later.

Michael C

Re: [Gdb] removing src/dejagnu, src/expect

[Christopher Faylor]

On Mon, Jun 07, 2004 at 05:57:57PM -0400, Michael Elizabeth Chastain wrote:
>Jim Ingham writes:
>>I just reran the gdb testsuite - on the Apple gdb but this shouldn't
>>make much difference - with the Expect 5.38 from the SourceForge
>>sources, and with Tcl 8.4.4, also from SourceForge, and I didn't see
>>any new failures compared with the expect & tcl from src/.
>
>I've done a similar comparison with the gdb testsuite on native
>i686-pc-linux-gnu, (red hat 8.0, LANG=en_US.UTF-8), and the results
>came out the same.  I've been using the released versions of
>tcl+expect+dejagnu for 2.5 years now.
>
>That's just gdb experience though.
>
>I'm okay with shooting first and fixing problems later.

The problem with that approach is that you can't necessarily fix all of
the problems can you?  In particular, as you noted in the gcc mailing
list, Cygwin could be a problem.  I'm not really looking to take on
new projects right now myself.  Maybe someone else will volunteer
to look at this as far as Cygwin is concerned.

cgf

Re: [Gdb] removing src/dejagnu, src/expect

[Michael Elizabeth Chastain]

> Well, please get over your mood and adopt a more proactive approach if
> you want to be test suite maintainer.  Don't be passive.  *Ask* if you
> want test suite results.  Cheesh.

Okay.  Chris, you build gdb on cygwin from time to time.  Next time
you're building gdb, can you run the test suite and post something to
gdb-testers?  Not now or any specific time, but whenever it's
convenient.

I haven't been aggressive about asking because I don't have the job of
test suite maintainer yet.

> FWIW, I don't object.  I don't care.

I don't object, either, but I do care.

Michael C

Re: [Gdb] removing src/dejagnu, src/expect

[Christopher Faylor]

On Tue, Jun 08, 2004 at 05:29:40PM -0400, Michael Elizabeth Chastain wrote:
>>The problem with that approach is that you can't necessarily fix all of
>>the problems can you?  In particular, as you noted in the gcc mailing
>>list, Cygwin could be a problem.
>
>You're right; I can't fix the problems.
>
>It would help if someone would post cygwin gdb results to gdb-testers
>so that we know if gdb even builds and the test suite runs.  The last
>cygwin results were from Raoul Gough on 2003-04-28.

Your faith in me is heartwarming.  Yes, gdb builds for cygwin, or at
least it did a month+ ago when I last checked.  I have many good reasons
for not being up-to-date on gdb right now but that is how free software
works, as you know.

>I've gotten into a mood where if gdb-testers doesn't contain any recent
>results for a platform, then I don't care too much if that platform
>breaks.  Which is actually a bad mood because that includes major
>platforms like solaris and cygwin.

Well, please get over your mood and adopt a more proactive approach if
you want to be test suite maintainer.  Don't be passive.  *Ask* if you
want test suite results.  Cheesh.

>Also please note that the decision to remove expect+dejagnu came from
>the Gcc Summit and was announced by Ben Elliston, who's going to do the
>actual removal on 2004-06-11, unless someone objects.

And, the gcc mailing list has been removed from this discussion.  Isn't
that conveeenient?

FWIW, I don't object.  I don't care.  When this goes away, I can
continue to maintain the cygwin fork of expect/dejagnu until such time
as "someone" steps forward to fix things.  The sources aren't going to
be physically deleted from src, so I can always access them.

Of course, it would also be nice if bje checked on any changes needed
for cygwin before pulling the plug.

cgf

Re: [Gdb] removing src/dejagnu, src/expect

[Michael Elizabeth Chastain]

> The problem with that approach is that you can't necessarily fix all of
> the problems can you?  In particular, as you noted in the gcc mailing
> list, Cygwin could be a problem.

You're right; I can't fix the problems.

It would help if someone would post cygwin gdb results to gdb-testers
so that we know if gdb even builds and the test suite runs.  The last
cygwin results were from Raoul Gough on 2003-04-28.

I've gotten into a mood where if gdb-testers doesn't contain any recent
results for a platform, then I don't care too much if that platform
breaks.  Which is actually a bad mood because that includes major
platforms like solaris and cygwin.

Also please note that the decision to remove expect+dejagnu came from
the Gcc Summit and was announced by Ben Elliston, who's going to do
the actual removal on 2004-06-11, unless someone objects.

Michael C

Re: [Gdb] removing src/dejagnu, src/expect

[Jim Ingham]

The gcc QA folks here apparently already use the built-in version of 
expect, not the src/ one for their automated testsuite runs.  They said 
they had to fix one 3.3 test, but that test is okay in 3.5.  So it 
sounds like we are okay.

The bit of info that needs to be transferred, then, is that whoever 
does the work to incorporate whatever patches we have to the expect 
sources shouldn't bother to try to get Ian's patches in, as they aren't 
really needed.

Jim

On Jun 8, 2004, at 5:26 AM, gdb-digest-help@sources.redhat.com wrote:

>
> Jim Ingham writes:
>> I just reran the gdb testsuite - on the Apple gdb but this shouldn't
>> make much difference - with the Expect 5.38 from the SourceForge
>> sources, and with Tcl 8.4.4, also from SourceForge, and I didn't see
>> any new failures compared with the expect & tcl from src/.
>
> I've done a similar comparison with the gdb testsuite on native
> i686-pc-linux-gnu, (red hat 8.0, LANG=en_US.UTF-8), and the results
> came out the same.  I've been using the released versions of
> tcl+expect+dejagnu for 2.5 years now.
>
> That's just gdb experience though.
>
> I'm okay with shooting first and fixing problems later.
>
> Michael C
>

Re: [Gdb] removing src/dejagnu, src/expect

[Jim Ingham]

Note, there are some local patches to Expect to switch the regular 
expression engine that Expect uses from the new Unicode aware version 
back to the original Henry Spencer one that shipped with Tcl 7.x.  A 
copy of the old Tcl regexp code was pulled into the src/ version of 
Expect.  This was done because the new regular expression engine was 
subtly different from the old one, this might cause spurious testsuite 
failures.  The ChangeLog for this is:

2001-07-09  Ian Roxborough  <irox@redhat.com>

         * tclParse-compat.c: New file.
         * tcl_regexp.c:
         * tcl_regexp.h: New file.  Contains Tcl8.0's regexp.
         * Makefile.in: Added new files to be compiled and linked.
         * exp_clib.c (exp_expectl, exp_fexpectl):
         * exp_inter.c (in_keymap, Exp_InteractCmd ):
         * exp_regexp.c (regtry, regdump, regprop):
         * expect.c (parse_expect_args, eval_case_string,
           exp_background_filehandler, Exp_ExpectCmd):
         * expect.h:
         * expect_comm.h:
         * Dbg.c (breakpoint_test, cmdBreak): Use newly built in
           regexp, not external Tcl regexp.
         * exp_tty.c (exec_stty):
         * exp_command.c (Exp_CloseCmd, Tcl_CloseCmd): Handle Tcl API
           changes in Tcl8.3.

I just reran the gdb testsuite - on the Apple gdb but this shouldn't 
make much difference - with the Expect 5.38 from the SourceForge 
sources, and with Tcl 8.4.4, also from SourceForge, and I didn't see 
any new failures compared with the expect & tcl from src/.  But this 
ought to be more broadly tested - and particularly with the gcc 
testsuite, before you do this switch.

Needless to say, the Expect folks are not very likely to take Ian's 
patch, since all it does is downgrades the regexp engine - sticking a 
second copy in the Expect sources...  So if there are new failures, 
then somebody is going to have to go through and clean up the 
testsuite.

Jim

On Jun 7, 2004, at 1:50 AM, gdb-digest-help@sources.redhat.com wrote:

>
>
> On Sat, Jun 05, 2004 at 10:39:42AM -0400, Michael Chastain wrote:
>>> At yesterday's GCC Summit testing BOF, there was a strong concensus
>>> that these directories should be removed from the src tree and that
>>> developers should install the latest releases of Expect and Dejagnu
>>> from their distribution/packaging system or from a source tarball at
>>> ftp.gnu.org.
>>
>> Err, I'm kinda surprised that the "GCC Summit" has enough src/
>> people to make decisions about src/.  Are Andrew Cagney (gdb)
>> and Daniel Jacobowitz (binutils) there?
>
> Yes (although I skipped the testing BOF out of sleepiness :).
>
> I'm redirecting this to the main GDB list.
>
> -- 
> Daniel Jacobowitz

Re: [Gdb] removing src/dejagnu, src/expect

[Daniel Jacobowitz]

On Sat, Jun 05, 2004 at 10:39:42AM -0400, Michael Chastain wrote:
> > At yesterday's GCC Summit testing BOF, there was a strong concensus
> > that these directories should be removed from the src tree and that
> > developers should install the latest releases of Expect and Dejagnu
> > from their distribution/packaging system or from a source tarball at
> > ftp.gnu.org.
> 
> Err, I'm kinda surprised that the "GCC Summit" has enough src/
> people to make decisions about src/.  Are Andrew Cagney (gdb)
> and Daniel Jacobowitz (binutils) there?

Yes (although I skipped the testing BOF out of sleepiness :).

I'm redirecting this to the main GDB list.

-- 
Daniel Jacobowitz