Open GoogleCodeExporter opened 8 years ago
Done.
Compiled r36-noasm3 using mingw32-3.4.4
One issue is that GCC misses empty base class optimisations sometimes, so some
changes will be needed to guard against it. For instance, global cache lookups.
The interpreter looks stable. Compiled code shows worse stability, i suppose
because
of wrong layout assumptions (empty base classes shifting the layout of all
subclasses)
Original comment by prunedt...@gmail.com
on 30 Apr 2007 at 2:17
Here is r36-noasm4, compiles under GCC 4.1 on debian etch, with the exception of
os_*.cpp files
Original comment by prunedt...@gmail.com
on 23 May 2007 at 10:32
Attachments:
Do the new crashes that occur happen in the compiler itself, the compiled code,
or
elsewhere in the VM? Although the compiler is the hardest part of the system
to work
on, it would be very nice to have a deterministic (script based) test that fails
under the new system but not under the old one. That might help a lot in
determining
the problem, since otherwise a new compiler bug might be very hard to isolate.
But
it might be a while before we have the expertise to do much hacking on the
compiler.
Original comment by David.Gr...@gmail.com
on 24 May 2007 at 8:46
Er, I can't download this file either. Prunedtree, could you please upload it
somewhere again?
Original comment by vincent....@gmail.com
on 14 Sep 2007 at 1:50
ok, it's alright now. I have managed to download it (after about 10 attempts).
Original comment by vincent....@gmail.com
on 14 Sep 2007 at 2:52
Original issue reported on code.google.com by
David.Gr...@gmail.com
on 22 Sep 2006 at 9:50