missing tests for 3.2.3-pre

missing tests for 3.2.3-pre

[Joe Buck]

The gcc-testresults archive does not show any test results for the
gcc 3.2.3 prerelease (of 20030416) for an Alpha platform, or for
FreeBSD.  The other primary platforms are covered.  The criteria
for 3.1, which we are still using for 3.2.x, specifies
i386-unknown-freebsd4.5 and alpha-unknown-linux-gnu.

Could anyone do tests?  I'd hate for us to ship 3.2.3 and have it
not work on BSD or Alpha.

Richard, you are listed as the Alpha tester; Loren, you are listed as
the FreeBSD tester.

Re: missing tests for 3.2.3-pre

[Joe Buck]

I wrote:
> The gcc-testresults archive does not show any test results for the
> gcc 3.2.3 prerelease (of 20030416) for an Alpha platform, or for
> FreeBSD.

We do have a test result from Matthias Klose for
3.2.3 20030415 (Debian prerelease) testsuite on alpha-unknown-linux-gnu 

However, this is for Debian's slightly modified version, so it's not
a test of Gaby's tarball.  But at least it indicates that we are in
good shape on the Alpha.

Re: missing tests for 3.2.3-pre

[Loren James Rittle]

In article <20030418110647.A7756@synopsys.com>,
Joe Buck<jbuck@synopsys.com> writes:

> The gcc-testresults archive does not show any test results for the
> gcc 3.2.3 prerelease (of 20030416) for an Alpha platform, or for
> FreeBSD.  The other primary platforms are covered.  The criteria
> for 3.1, which we are still using for 3.2.x, specifies
> i386-unknown-freebsd4.5 and alpha-unknown-linux-gnu.

> Could anyone do tests?  I'd hate for us to ship 3.2.3 and have it
> not work on BSD or Alpha.

> Richard, you are listed as the Alpha tester; Loren, you are listed as
> the FreeBSD tester.

Hi Joe, thank you for taking an active interesting in monitoring
release quality.  I've started the process for pre2.  I'm sorry, but
these things take some time when there are *three* active branches of
development.  A detailed report on the status of pre1 went directly to
the gcc 3.2.3 release manager within 3 days of pre1's release.  I will
include it below since I didn't CC the gcc list.  Summary, there are
no interesting regressions on: i386-unknown-freebsd4.8 (-STABLE),
i386-unknown-freebsd5.0 (-CURRENT), alpha-unknown-freebsd5.0
(-CURRENT).  Also, all current failures are either bogus, caused by
FreeBSD-side changes (and more-or-less harmless) or well-known for
this port.  I don't actually have a freebsd4.5 machine around to test
anymore.  I'm currently expecting a similar total cycle time for pre2.

Regards,
Loren

Date: Tue, 15 Apr 2003 06:50:08 -0500 (CDT)
From: Loren James Rittle <rittle@latour.rsch.comm.mot.com>
Message-Id: <200304151150.h3FBo8aF077319@latour.rsch.comm.mot.com>
To: gdr@integrable-solutions.net
CC: pfeifer@dbai.tuwien.ac.at
Subject: Notice to 3.2.3 release manager (a.k.a Gaby)

Hi Gaby,

First of all, thanks for doing the hard job of release manager of FSF
gcc 3.2.3.  Here is a primary evaluation platform report (only the
FreeBSD4/i386 port is so listed, but it was just as easy to do all of
them concurrently).  Status: functionality-wise we are good-to-go on
this platform.  Tested against dejagnu framework before install only,
thus I didn't test any aspects of performance or real-world tests.
Tested 'gmake install' and then spot-checked the installed compiler.

I have bootstrapped the 04-12-2003 tarball (all languages, save ada)
which you put up for testing and analyzed the test results on:

i386-unknown-freebsd4.8 "with objdir configuration, system binutils, GNU make"

    http://gcc.gnu.org/ml/gcc-testresults/2003-04/msg00891.html

    No regressions; zero real failures, save libjava issues.

i386-unknown-freebsd5.0 "with srcdir configuration, system binutils, GNU make"
alpha-unknown-freebsd5.0 "with objdir configuration, system binutils, BSD make"

    http://gcc.gnu.org/ml/gcc-testresults/2003-04/msg00888.html

    All of these regressions appear related to the fact that C99
    support is now enabled when libstdc++-v3 is configured on this
    port (due to libc upgrade, not any change in libstdc++-v3):

    +FAIL: 26_numerics/c99_classification_macros_c.cc (test for excess errors)
    +FAIL: 27_io/ostream_inserter_arith.cc execution test
    +FAIL: "all g++ test cases compiled with -ansi"

    Mostly harmless.  No patch will be forthcoming for the 3.2.X branch.

  These are all bogus FAILs seen on all three platforms (related to
  the bison version used to build the files in the tarball?):

  +FAIL: objc.dg/naming-1.m  (test for errors, line 20)
  +FAIL: objc.dg/naming-1.m (test for excess errors)
  +FAIL: objc.dg/naming-2.m  (test for errors, line 7)
  +FAIL: objc.dg/naming-2.m (test for excess errors)

Regards,
Loren

Re: missing tests for 3.2.3-pre

[Joe Buck]

I wrote: 
> > The gcc-testresults archive does not show any test results for the
> > gcc 3.2.3 prerelease (of 20030416) for an Alpha platform, or for
> > FreeBSD.  The other primary platforms are covered.  The criteria
> > for 3.1, which we are still using for 3.2.x, specifies
> > i386-unknown-freebsd4.5 and alpha-unknown-linux-gnu.
 
On Fri, Apr 18, 2003 at 04:36:52PM -0500, Loren James Rittle wrote:
> Hi Joe, thank you for taking an active interesting in monitoring
> release quality.

I was basically asked to be the "quality guy" way back when we started
egcs, though I haven't always had the cycles to do a great job at it.

>  I've started the process for pre2.  I'm sorry, but
> these things take some time when there are *three* active branches of
> development.  A detailed report on the status of pre1 went directly to
> the gcc 3.2.3 release manager within 3 days of pre1's release.  I will
> include it below since I didn't CC the gcc list.  Summary, there are
> no interesting regressions on: i386-unknown-freebsd4.8 (-STABLE),
> i386-unknown-freebsd5.0 (-CURRENT), alpha-unknown-freebsd5.0
> (-CURRENT).

> ...  I don't actually have a freebsd4.5 machine around to test
> anymore.  I'm currently expecting a similar total cycle time for pre2.

I think that it would be fine to treat 4.8 as the primary reference
platform, since it's the current stable release.  The primary platform
list says 4.5, but it was done back before the 3.1 release, so it's
out of date.

Notice to 3.2.3 release manager (a.k.a Gaby), RE pre2

[Loren James Rittle]

Hi Joe,

> I think that it would be fine to treat 4.8 as the primary reference
> platform, since it's the current stable release.  The primary platform
> list says 4.5, but it was done back before the 3.1 release, so it's
> out of date.

I concur.  I have commited a patch to the 3.3 primary platform
criteria page (there was none for 3.2).  Instead of naming an exact
numeric release, I wrote "-STABLE/-CURRENT" since that is what
actually gets tested every day.

i386-unknown-freebsd4.8 has now completed bootstrap/check/install of pre2.
alpha-unknown-freebsd5.0 has now completed bootstrap/check/install of pre2.

Hi Gaby,

>>  I've started the process for pre2.  [...]

Here is a primary platform report (only the FreeBSD4/i386 port is so
listed but it was easy to do FreeBSD/alpha at the same time).  Status:
functionality-wise we are good-to-go on this platform.  Tested against
dejagnu framework before install only, thus I didn't test any aspects
of performance or real-world tests.  Tested 'gmake install' and then
spot-checked the installed compiler.

I have bootstrapped the 04-16-2003 tarball (all languages, save ada)
which you put up for testing and analyzed the test results on:

i386-unknown-freebsd4.8 "with srcdir configuration, system binutils, GNU make"

    http://gcc.gnu.org/ml/gcc-testresults/2003-01/msg01313.html (3.2.2)
    http://gcc.gnu.org/ml/gcc-testresults/2003-04/msg00891.html (pre1)
    http://gcc.gnu.org/ml/gcc-testresults/2003-04/msg01179.html (pre2)

    No regressions; zero real failures, save documented libjava issues.

alpha-unknown-freebsd5.0 "with objdir configuration, system binutils, GNU make"

    http://gcc.gnu.org/ml/gcc-testresults/2003-01/msg01311.html (3.2.2)
    http://gcc.gnu.org/ml/gcc-testresults/2003-04/msg00888.html (pre1)
    http://gcc.gnu.org/ml/gcc-testresults/2003-04/msg01177.html (pre2)

    All of these regressions appear related to the fact that C99
    support is now enabled when libstdc++-v3 is configured on this
    port (due to libc upgrade, not any change in libstdc++-v3):

    +FAIL: 26_numerics/c99_classification_macros_c.cc (test for excess errors)
    +FAIL: 27_io/ostream_inserter_arith.cc execution test
    +FAIL: "all g++ test cases compiled with -ansi"

    Mostly harmless.  Due to timing, no patch will be forthcoming for
    the 3.2.X branch (although I think I finally understand how
    to create a near-zero risk patch, my track record for breaking...)

Regards,
Loren

Re: missing tests for 3.2.3-pre

[Gabriel Dos Reis]

Joe Buck <jbuck@synopsys.com> writes:

[...]

| Richard, you are listed as the Alpha tester; Loren, you are listed as
| the FreeBSD tester.

Hi Joe,

I got very detailed reports from Loren in private.  
I see I should have made the list aware of that.  
I'm going to send a report on the current status and see what is missing.

Thanks,

-- Gaby