Re: release 3.0.2

[Jaroslav Hajek]

On Tue, Aug 19, 2008 at 2:25 PM, Sergei Steshenko <address@hidden> wrote:
>
>
>
> --- On Tue, 8/19/08, Jaroslav Hajek <address@hidden> wrote:
>
>> From: Jaroslav Hajek <address@hidden>
>> Subject: Re: release 3.0.2
>> To: address@hidden
>> Cc: "Thomas Weber" <address@hidden>, address@hidden
>> Date: Tuesday, August 19, 2008, 5:16 AM
>> On Tue, Aug 19, 2008 at 2:15 PM, Sergei Steshenko
>> <address@hidden> wrote:
>> >
>> >
>> >
>> > --- On Tue, 8/19/08, Jaroslav Hajek
>> <address@hidden> wrote:
>> >
>> >> From: Jaroslav Hajek <address@hidden>
>> >> Subject: Re: release 3.0.2
>> >> To: "Thomas Weber"
>> <address@hidden>
>> >> Cc: address@hidden
>> >> Date: Tuesday, August 19, 2008, 1:12 AM
>> >> > I would also prefer to have the tarballs on
>> the normal
>> >> Octave page.
>> >>
>> >> It's going to happen, but I have no access.
>> John is
>> >> just waiting
>> >> whether all the regular builders can build this
>> release. I
>> >> pre-announced the release two weeks ago, but
>> perhaps I
>> >> should have
>> >> made a second "last call".
>> >> Perhaps I could do with a list of regular
>> builders. I would
>> >> wait for
>> >> all of them to report a successful build from the
>> archive,
>> >> and then
>> >> make the official tarballs.
>> >>
>> >>
>> >
>> > Well, not quite successful.
>> >
>> > The build is OK, but there is a failures in 'make
>> check':
>> >
>> >     18   src/DLD-FUNCTIONS/rand.cc
>> .............................. PASS   56/57   FAIL 1
>> >
>> > Regards,
>> >  Sergei.
>> >
>>
>> I know about this one. An extra bracket. I don't think
>> this is worth
>> remaking the release.
>>
>> thanks,
>>
>> --
>> RNDr. Jaroslav Hajek
>> computing expert
>> Aeronautical Research and Test Institute (VZLU)
>> Prague, Czech Republic
>> url: www.highegg.matfyz.cz
>
> I do not think it is decent to make a release of whatever SW product if
> it is known that built-in test suite fails.
>

Well it happens a lot, doesn't it? I recall a bunch failures when I
first tried to compile Octave 3.0.1 with Intel C++. Admittedly, this
test will fail always, but I don't see this as a serious regression.

> It looks like the fix is really simple, so why not either to undo the
> release or to make 3.0.2.a ?
>

Most patches are simple to apply. I think this just comes down to the
question what problems should be considered serious enough to
unfreeze.

Currently, I think of these:

1. Build failures
2. Segfaults during make check.
3. Serious test failures during make check.
Preferably, this should all be carried out stable builders, who are
most likely to have everything configured correctly.

By "serious" test failure I mean something is detected that is bound
to cause problems. For instance, there have been tests in the suite
that were simply failing because the error threshold was set too low
for some configurations. That's not a serious test failure. I don't
see a typo in a test as a serious failure either.

So we seem to be in dispute over whether 3 should read "any test
failure...". I count your vote, but I still disagree. What do others
think?

> I'll rebuild - for me it's just to change version and wait.
>
> Thanks,
>  Sergei.
>
>
>
>



-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz