Re: Jumbo libgcj patch

[Bryce McKinlay <bryce at albatross dot co dot nz>]

Tom Tromey wrote:

> * Adds a first cut at serialization.  This is a somewhat modified
>   version of the Classpath serialization code.  It is known not to
>   correctly interoperate with other serialization implementations yet.
>   (The interoperability and divergence-from-Classpath problems will be
>   addressed.)
>

While I confess to not knowing too much about serialization, wouldn't
"interoperability with other implementations" mean that the standard libraries
would have to share the exact same names and uses for all their non-transient
private fields? Is that a realistic goal? Or do you mean interoperability at the
basic protocol level?

> I'm not going to post the patch since it is huge.  I've appended the
> ChangeLog entry.
>
> I realize it would be better to do this as a series of independent
> patches.  That is too much work though, and I don't have the time :-(.

This changelog entry is duplicated:

2000-03-26  Hans Boehm  <boehm@acm.org>

 * include/java-cpool.h (_Jv_storeLong, _Jv_loadLong,
 _Jv_storeDouble, _Jv_loadDouble): Define differently on 64 bit
 machine.
 * java/lang/ieeefp.h: Define __IEEE_BIG_ENDIAN or
 __IEEE_LITTLE_ENDIAN appropriately on IA64.
 * java/lang/mprec.h: Don't define Pack_32 on 64 bit machine.
 * javaprims.h (_Jv_word): Added `l' and `d' entries in 64 bit
 case.
 * resolve.cc (FFI_PREP_RAW_CLOSURE): New define.
 (FFI_RAW_SIZE): Likewise.
 (_Jv_InterpMethod::ncode): Use them.
 * interpret.cc (PUSHL, PUSHD, POPL, POPD, LOADL, LOADD, STOREL,
 STORED): Define differently on a 64 bit machine.
 (continue1): Use ffi_java_raw_call when appropriate.

regards

  [ bryce ]