OpenJDK zero on sparc linux

[Andrew Haley]

Gary Benson wrote:
> Andrew Haley wrote:
>> This isn't really a bug; we just need someone with access to
>> Sun hardware to write __sync_val_compare_and_swap_4,
>> __sync_lock_test_and_set_4, and __sync_add_and_fetch_4.
> 
> Speaking of which... Andrew, Xerces posted a patch with arm-specific
> versions of these functions.  Would it be possible to get these rolled
> into gcc rather than committing them as workarounds in icedtea?

No, because gcc only generates these when the processor has hardware
support; the RAM has nonse, so Xerces' code will only work on Linux.

It really is hard to do this well and portably, so we're going to have
to live with this on targets that don't have support.

For a portable workaround, we can perhaps do something with pthreads
mutexes.  Not nice, but it'd work.

Andrew.