1.6 changes to make ARM work with gcc 4.1

[Andrew John Hughes]

2009/8/26 Edward Nevill <ed at camswl.com>:
> Hi all,
>
> Below is a patch which will make the ARM interpreter build with gcc-4.1
> series.
>
> I am not sure whether this should go in the 1.6 release or not. On the
> one hand it is a shame it does not build with 4.1, on the other hand
> very few people build with 4.1.
>
> The basic change is to inline those functions which were causing unresolved
> symbols on gcc 4.1. The reason the symbols were unresolved was that
> gcc 4.1 inlined these functions whereas gcc 4.4 did not. Therefore
> when the assembler tried to call these functions they were unresolved
> because they had been inlined.
>
> Later vsns of gcc seem to get around this by putting the inlined functions
> in a .comdat section. This is then excluded by ld if it is not referenced.
>
> The inlining I have done is pretty much a cut and paste from the gcc 4.4
> output.
>
> Note that since it is now no longer dependant on the GCC_VERSION I have
> removed that.
>
> I have also reduced the code and data alignment from 64 byte alignment
> to 8 byte alignment as earlier binutils had problems with 64 byte alignment.
> This was only a very minor performance optimisation anyway.
>
> As I say, I am in two minds whether this should go in 1.6 or not. Your
> opinions please.
>
> Regards,
> Ed.
>

Does it cause any regressions or have a noticeable effect on users
using gcc > 4.1? If not, it's a sensible fix to have in the release
which will make it usable by more people.
-- 
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