Re: [patch] Workaround ada (gnat-4.3.0) on x86_64

[Joel Brobecker <brobecker at adacore dot com>]

> Does the testcase work for you with GDB HEAD gdb.ada/packed_tagged.exp?
> The same memory trashing problem is there for me for gcc-4.1.2-33.i386
> and for gcc-4.3.0-8.i386 (Fedora 8 and 9, resp.) when trying the ia32
> arch.

Yes, these testcases work for me with GDB HEAD; but I'm using GNAT Pro,
which produces the expected debugging info.

Obviously, there are some local changes that AdaCore hasn't contributed
yet. I know that some of them were quite tricky to get right, so we
elected to keep them local while we build confidence in them.  But
both Eric Botcazou and Olivier Hainque would like to work towards
contributing them (and I know that Olivier improved some of them
towards that goal). This is something I wanted to discuss more in
depth with them at the GCC summit - anyone attending who would like
to know more about that task is welcome to discuss it with us.

> While it does not fix the FAIL it has no regressions (for the gcc testsuite on
> i386+x86_64 and with the GDB testsuite) and it no longer trashes the memory
> by omitting the unknown DW_AT_byte_size values (instead of writing them out
> with the value 0xffffffff):
> 	GCC PR 35998 http://gcc.gnu.org/bugzilla/attachment.cgi?id=15654

Unfortunately, I am not the right person to discuss compiler changes,
as I very seldomly modified the GCC. I'll ask Eric what he thinks
about it.

-- 
Joel