This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [4.0 PATCH] Disable Tree-SRA for the Ada compiler
On Feb 26, 2005, at 05:52, Eric Botcazou wrote:
I think that's premature. And unfortunate since SRA helps Ada code a
lot.
But, as far as I can tell, it will introduce a brand new class of bugs
too, at
the interfaces between types and subtypes. See PR ada/18819.
I feel that for now, it is premature to turn of SRA on the branch,
even though I think we'll probably end up doing just that for the final
4.0 release.
The reason for keeping SRA is that most of the bugs found with SRA
would be latent bugs that might also be triggered in different
circumstances.
So, as long as we're in the mode of fixing these bugs on both mainline
and branch, we should keep SRA active.
Once we get at the point that either potentially destabilizing changes
outside the Ada front end are necessary, or mainline and branch diverge
too much for bugfixes to be applicable to both, then I think we should
turn off SRA for Ada.
If we disable it, we're likely to end up keeping it that way.
On the 4.0 branch, sure. But since it will not be disabled on
mainline, I
don't really see the problem.
I think leaving SRA enabled for now on the branch, we may end up
with a better compiler.
-Geert