6b18 build problems

[Andrew John Hughes]

On 20 February 2010 14:35, Matthias Klose <doko at ubuntu.com> wrote:
> On 20.02.2010 12:19, Andrew John Hughes wrote:
>>
>> On 19 February 2010 23:28, Matthias Klose<doko at ubuntu.com>  wrote:
>>>
>>> seen some build problems with 6b18
>>>
>>
>>  From the rest of your email it sounds like you actually mean IcedTea6
>> with the recent updates, rather than upstream OpenJDK6 in most cases.
>
> yes (how much is "most" for three cases ?-)
>

All but the first seem to be related to IcedTea6, but my point was
more that I wanted to know whether you were doing an IcedTea build
(./configure && make) or an  OpenJDK build (make <numerous_options>)..
 I regularly do both, so it's not immediately obvious to me which one
is being discussed if you just mention an OpenJDK tarball.

>>>  - ant-1.8 returns an error code != 0 calling ant -diagnostics,
>>>   breaking the build.
>>
>> Haven't seen this new version yet, but I would guess that is an
>> OpenJDK or Ant issue.
>>
>> if this is intended, then we need a
>>   workaround in openjdk.
>
> it looks like the build now requires the replaceregexp task, which at least
> in Debian is in the ant-optional package. However the exit code of ant
> -diagnostics doesn't change with this package installed.
>

Yes, the replaceregexp task has been needed for a while for 7.  It's
to do with the change to drop-in tarballs for JAXP and JAXWS.

I've not got a system with such a new Ant yet (is this some unstable
Ubuntu you're testing on?) so not currently in a situation to see this
bug.

>>>  - cacao build failure when configured with
>>>     --with-openjdk=/usr/lib/jvm/java-6-openjdk
>>>     --with-hotspot-build=original
>>>     --with-additional-vms=cacao
>>>   works with a two stage build using gcj.
>>>
>>
>> --with-cacao was broken until yesterday due to one of the patches.  I
>> haven't seen this failure and it's odd if it only occurs when using
>> --with-openjdk and not a full build.  Why would java.nio.ByteBuffer be
>> missing?  It's even in GCJ.
>
> the build did include these changes.
>

What changes?  It does make some bootstrapping changes because these
are necessary with the JAXP and JAXWS sources vanishing.  I don't see
how this would affect ByteBuffer and certainly not on an OpenJDK build
which doesn't need the bootstrapping anyway.

I'll have a look when I'm back at work on Monday.

>>>   details:
>>>
>>>
>>> http://launchpadlibrarian.net/39430688/buildlog_ubuntu-lucid-armel.openjdk-6_6b18~pre1-0ubuntu1_FAILEDTOBUILD.txt.gz
>>>
>>>  - make jtregcheck starts rebuilding the whole openjdk again. apparently
>>>   an out of date stamp, or a missing stamp. still searching for this one.
>>>   any ideas?
>>
>> No, other than make sure you're using the latest HEAD as I did fix a
>> few issues yesterday.
>
> no, didn't fix. Mark (W.), is this seen on the buildd's as well, when
> running make && make check?
>

If you're referring to Mark's builds on the icedtea.classpath.org
servers, they don't seem to be getting as far as make check for some
reason I can't yet fathom.

We have automated tests going on a box inside Red Hat   I haven't yet
checked that in detail yet to see if it is repeating part of the
build.  Probably some dead Makefile target.  I think it's probably
worth documenting the current targets somewhere and nuking any that
are unused/unneeded and more importantly not well-maintained.

This is just thoughts off the top of my head; I'll look into the
issues on Monday.

>  Matthias
>

Cheers,
-- 
Andrew :-)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

Support Free Java!
Contribute to GNU Classpath and the OpenJDK
http://www.gnu.org/software/classpath
http://openjdk.java.net

PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8