This is the mail archive of the
mailing list for the GCC project.
Re: Limit cselib usage
> > I am somewhat confused by the history log of 14462. Do you have some
> > kind of self contained testcase (except of broken ADA on hppa I can not
> > test?) or some other clue that can let me figure out what is
> > wrong?
> The log got a bit off topic because there were two separate breaks
> of the Ada bootstrap. The first was fixed several days latter by
> Richard Kenner's patch to trans.c. I spent quite a bit of time
> trying to track down the cause of the first break only to find that
> Richard had already fixed it ;(
> As for a testcase, I don't have one. My knowledge of Ada is limited.
My too, making things more dificult :(
In general the patch is just makind CSElib to have less matches in the
cases the matches would be useless anyway. It is kind of surprising that
it causes code to be miscompilled...
> As for testing on hppa, arrangements can be made. I believe an
> account was setup for you on gsyprf11.external.hp.com a few months
Ah, I forgot about this entirely. I will check it out.
> > I can revert the patch (do I need approval for this?) but that won't
> > help to fix it and I can't look on it myself.
> For the reasons listed in your patch submission, it would be nice if
> this could get fixed. However, I'm away for 8 days next week and my
> time is very limited for awhile. PR 13324 appears similar and involved
> the movstrsi/clrstrsi patterns.
Another one with long history. I will try to look into it and figure
out if there are some other fialures attributed to this patch.
> J. David Anglin email@example.com
> National Research Council of Canada (613) 990-0752 (FAX: 952-6602)