--- Comment #13 from matz at gcc dot gnu dot org 2007-11-01 03:11 ---
This one is fixed now.
--
matz at gcc dot gnu dot org changed:
What|Removed |Added
Statu
--- Comment #12 from matz at gcc dot gnu dot org 2007-11-01 03:06 ---
Subject: Bug 33961
Author: matz
Date: Thu Nov 1 03:06:38 2007
New Revision: 129817
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=129817
Log:
PR tree-optimization/33961
* tree-ssa-phiopt.c (st
--- Comment #11 from matz at gcc dot gnu dot org 2007-10-31 23:12 ---
Your threading testcase also seems to work with my patch. It doesn't fail
really reliably for me with the other compilers, so I'm not 100% certain,
can you check? Patch at
http://gcc.gnu.org/ml/gcc-patches/2007-10
--- Comment #10 from jakub at gcc dot gnu dot org 2007-10-31 22:45 ---
Thanks. Please see also PR31862.
Looking at 456.hmmer it seems at least in the routines I have looked at there
is always some unconditional write within the loop to the conditionally stored
variable, so perhaps the b
--- Comment #9 from matz at gcc dot gnu dot org 2007-10-31 21:52 ---
Working on it.
--
matz at gcc dot gnu dot org changed:
What|Removed |Added
AssignedTo|unassig
--- Comment #8 from jakub at gcc dot gnu dot org 2007-10-31 19:25 ---
This is valid code and cselim really has to be changed not to introduce writes
to possibly shared or read-only variables that are otherwise never written
into.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33961
--- Comment #7 from bero at arklinux dot org 2007-10-31 17:54 ---
To clarify, the problem in this usage scenario occurs because the test case
crashes even though the code path that writes to the read-only memory is never
actually run (there is no \0 in the string).
if(d[i] == 0) {
c
--- Comment #6 from maksim at kde dot org 2007-10-31 17:45 ---
The usage scenario of the method pretty much precludes it from being used in a
way where read-only memory would have to be written, except if someone does
something silly with the API. Essentially it is called either with out
--- Comment #5 from pinskia at gcc dot gnu dot org 2007-10-31 17:07 ---
So what we have is a conditional store but a non conditional read. So the
store can trap but the read does not. Really using a const_cast does nothing
to change the fact you are passing a readonly memory to somethi
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
Severity|blocker |normal
Component|rtl-optimization|tree-optimizat
--- Comment #3 from bero at arklinux dot org 2007-10-31 17:00 ---
That conversion warning is only in the simplified test case, the original code
it's extracted from (khtml) doesn't pass a string constant (it passes html code
retrieved from a web server) and crashes the same way.
There i
--- Comment #2 from pinskia at gcc dot gnu dot org 2007-10-31 16:53 ---
This is the new pass by Matz which is causing the issue. So we have a
conditional write which writes to a read only memory.
Though the warning:
vt.cc: In function 'int main(int, char**)':
t.cc:21: warning: deprecat
12 matches
Mail list logo
