This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: tree-ssa branch: Call for preliminary testing
- From: Andrew MacLeod <amacleod at redhat dot com>
- To: Daniel Berlin <dberlin at dberlin dot org>
- Cc: Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>,Diego Novillo <dnovillo at redhat dot com>,Steven Bosscher <s dot bosscher at student dot tudelft dot nl>,gcc mailing list <gcc at gcc dot gnu dot org>
- Date: 12 May 2003 10:16:26 -0400
- Subject: Re: tree-ssa branch: Call for preliminary testing
- References: <C80E3277-8482-11D7-87A1-000A95A34564@dberlin.org>
On Mon, 2003-05-12 at 10:05, Daniel Berlin wrote:
>
> On Monday, May 12, 2003, at 09:48 AM, Richard Guenther wrote:
>
> > On 12 May 2003, Diego Novillo wrote:
> >
> >> On Mon, 2003-05-12 at 08:56, Steven Bosscher wrote:
> >>
> >>> POOMA even has its own page in the GCC wwwdocs!
> >>> Ref. http://gcc.gnu.org/testing/testing-pooma.html
> >>>
> >> Shame on me. Dobby will go iron his hands now.
> >
> > Got something more interesting...
> >
>
> This abort is a known thing (it's because of an unfinished piece of
> out-of-SSA), though I swear Diego was under the impression it shouldn't
> happen with any of the current passes.
>
It usually happens now only when we've done something wrong. If we have
deleted a definition without removing all the uses, the live range
analyzer will make that variable live all the way from its use to the
beginning of the program, considering it to be live on entry to the
function.
THis will usually cause it to interfere with some other version of that
variable, and then this abort triggers.
Put a breakpoint in the abort(),
p print_genric_expr (stderr, t, 2)
and this will be the variable for which you are probably can't find a
DEF for. Figure out who deleted it and why, and thats likely your bug
Andrew.