This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: GCC 3.4.1 RC1 available
- From: tbp <gcc at ompf dot org>
- To: gcc at gcc dot gnu dot org
- Date: Sun, 27 Jun 2004 05:11:37 +0200
- Subject: Re: GCC 3.4.1 RC1 available
- References: <40DDCBF4.9090200@codesourcery.com>
Mark Mitchell wrote:
The first release candidate for GCC 3.4.1 is available here:
ftp://gcc.gnu.org/pub/gcc/prerelease-3.4.1-20040625/
The goal of release candidates is partly to obtain additional testing --
Here you are: built on cygwin; seems ok.
Tho i've hit something:
/usr/local/gcc-341/bin/g++.exe -DNDEBUG -DWIN32 -Id:/src/tbp/inc
-Id:/src/tbp/lib -Id:/incs -gstabs+ -pipe -Wall -Wdisabled-optimization
-Wno-unused -Wformat=2 -fmessage-length=0 -ffast-math -fstrict-aliasing
-fomit-frame-pointer -mfpmath=sse -march=k8 -O2 -malign-stringops
-minline-all-stringops -ftracer -funit-at-a-time -fpeel-loops
-funroll-loops -funswitch-loops -mwindows -mcygwin -fno-rtti
-fno-exceptions -fno-implement-inlines -fno-enforce-eh-specs
-fno-implicit-templates -mcygwin
-Wp,-MT,build/rtrace.o,-MMD,build/rtrace.d -c src/rtrace.cpp -o
build/rtrace.o
src\rtrace.cpp(1): warning: `build/rtrace.gcda' is version `
Internal compiler error: Error reporting routines re-entered.
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http:\\gcc.gnu.org\bugs.html> for instructions.
That rtrace.gcda file was from a previous 3.4.0 pgo run and while i've
never told gcc to use it on that compile it read it and choked to death.
A symptomatic cure was to remove the offending files.
Dunno if it's a known bug.
Note that proper pgo runs seem ok but i can't really say because as
usual on that platform when using SSE the final bin blows and i can't be
arsed to check that particular bin and do the usual combination of ad
hoc stack fixing and/or symbol/constant hunting etc...
[ yes, i'm ranting a bit but while i really like to use gcc and once you
kludge around those issues it's way faster than the competition, i'm
wasting hours fixing things each time i dare to change a compilation
switch ]
Also i haven't checked yet if the codegen problem related to
_mm_load/store_si128 is still there.