This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Eliminating Assembler Already Defined Messages for Java Library


Things seems to be working better now - at least compiling
gnu.bytecode.dump produced a working executable.  I will see if I
can check in my patches tonight.

My previous patches were all "wrong", in the sense that my current
sources are completely different.  It turns out that jc1 generate
*two* blocks for each source block: The parser (first pass) uses
enter_block/exit_block, which creates BLOCK object on each
enter_block.  Then, after name resolution, we expand the result,
calling pushlevel/ poplevel for each BLOCK we got from the first pass;
each poplevel creates a *new* BLOCK.  Thus setting
x_whole_function_mode_p flag is probably the wrong thing, given the
way things are done.  On the other hand, generating double sets of
BLOCKs is perhaps a bit screwey, not to say confusing, but I don't
want to fix that now.  When I do feel the urge, I think it might be
reasonable to leave the expend-time pushlevel/poplevels in place, and
have the parser create BIND_EXPR objects instead of BLOCK objects.

Anyway, I will be sending out messages on various patches as I check
them in (to the trunk).  Some of the patches are unrelated, and some
are just general tidying things up, but it is worth getting them all
in, to sync my tree with the cvs tree.

After that, I will try to merge my changes into the gcc3 branch,
as I am hoping that people will be able to build Kawa using gcc3.
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/~per/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]