openjdk6b11 built for linux-armeabi using zero

[Matthias Klose]

this is gcc-4.3, taken from the current gcc-4_3-branch, with r136501 backed out
(the reversion is only important for hotspot builds). libjava has a patch taken
from classpath trunk to make gjar understand the @<file> syntax. This patch can
be found here:
http://svn.debian.org/viewsvn/*checkout*/gcccvs/branches/sid/gcc-4.3/debian/patches/libjava-file-support.dpatch?rev=3163
Note that you have to add the class files as well.

binutils is taken from the 2.18 branch.

  Matthias

Xerxes Rånby schrieb:
> Fantastic news!
> 
> I have myself tried to build icedtea6 for armv5tel eabi for some weeks
> now using a fedora core 8 based buildsystem.
> I have had a lot issues that my GCC 4.1.2 compiler mangles pointers with
> the -O3 optimization when starting the launcher.
> Those errors disapeared with -O0 yet then i ran out of heap space when
> launching. Adding options to java like -Xms100M -Xmx600M did not solve
> my heapspace issues.
> 
> What version of gcc compiler and binutils did you use for your build?
> 
> Congratulations again for the sucessful ARM build!
> 
> Have a great day!
> Xerxes
> 
> Matthias Klose skrev:
>> See http://people.ubuntu.com/~doko/java/openjdk6-armel/
>>
>> - the mauve test results are a subset of mauve tests which is run
>>   on various buildds. I usually get 4 failures for the hotspot
>>   archs, and 6 for powerpc zero. I don't run the jtreg test harness,
>>   this probably would take two weeks on this machine.
>>
>> - the packages should appear with the next upload to unstable, well ...
>>   some days or weeks after the source upload, depending on the build
>>   machine. the native build takes ages, the longest step is the
>>   jar process building the rt.jar in the stage2 build; took over 1200
>>   cpu minutes on a 500Mhz 256MB machine.
>>
>> Matthias
>>   
>