From: SourceForge.net <no...@so...> - 2004-04-23 19:45:13
|
Bugs item #940940, was opened at 2004-04-23 15:45 Message generated for change (Tracker Item Submitted) made by Item Submitter You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=102435&aid=940940&group_id=2435 Category: gcc Group: Known bugs Status: Open Resolution: None Priority: 5 Submitted By: Doug Currie (dougcurrie) Assigned to: Danny Smith (dannysmith) Summary: optimization error building zlib1.2.1 Initial Comment: With gcc.exe (GCC) 3.3.3 (mingw special) compiling the zlib 1.2.1 distribution from http://www.gzip.org/zlib/ with the zlib supplied win32/Makefile.gcc the built-in tests fail. The problem occurs in the function int ZEXPORT gzclose (file) wherein the value read for s->mode is incorrect. If in win32/Makefile.gcc I replace CFLAGS = $(LOC) -O3 -Wall with CFLAGS = $(LOC) -O2 -Wall the tests pass. I conclude that gcc.exe (GCC) 3.3.3 (mingw special) has an optimization related bug. Other configuration info: Windows XP sp1 e@TRUFFLE ~ $ gcc -v Reading specs from c:/msys/mingw/bin/../lib/gcc- lib/mingw32/3.3.3/specs Configured with: ../gcc/configure --with-gcc --with-gnu- ld --with-gnu-as --host=mingw32 --target=mingw32 -- prefix=/mingw --enable-threads --disable-nls --enable- languages=c,c++,f77,ada,objc,java --disable-win32- registry --disable-shared --enable-sjlj-exceptions -- enable-libgcj --disable-java-awt --without-x --enable- java-gc=boehm --disable-libgcj-debug --enable- interpreter --enable-hash-synchronization Thread model: win32 gcc version 3.3.3 (mingw special) e@TRUFFLE ~ $ ld -v GNU ld version 2.15.90 20040222 e@TRUFFLE ~ $ uname -a MINGW32_NT-5.1 TRUFFLE 1.0.10(0.46/3/2) 2004-03- 010 08:07 i686 unknown installed from gcc-core-3.3.3-20040217-1.tar.gz gcc-g++-3.3.3-20040217-1.tar.gz binutils-2.15.90-20040222-1.tar.gz MSYS-1.0.10.exe e ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=102435&aid=940940&group_id=2435 |
From: SourceForge.net <no...@so...> - 2004-04-25 05:31:41
|
Bugs item #940940, was opened at 2004-04-24 07:45 Message generated for change (Comment added) made by dannysmith You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=102435&aid=940940&group_id=2435 Category: gcc Group: Known bugs Status: Open Resolution: None Priority: 5 Submitted By: Doug Currie (dougcurrie) Assigned to: Danny Smith (dannysmith) Summary: optimization error building zlib1.2.1 Initial Comment: With gcc.exe (GCC) 3.3.3 (mingw special) compiling the zlib 1.2.1 distribution from http://www.gzip.org/zlib/ with the zlib supplied win32/Makefile.gcc the built-in tests fail. The problem occurs in the function int ZEXPORT gzclose (file) wherein the value read for s->mode is incorrect. If in win32/Makefile.gcc I replace CFLAGS = $(LOC) -O3 -Wall with CFLAGS = $(LOC) -O2 -Wall the tests pass. I conclude that gcc.exe (GCC) 3.3.3 (mingw special) has an optimization related bug. Other configuration info: Windows XP sp1 e@TRUFFLE ~ $ gcc -v Reading specs from c:/msys/mingw/bin/../lib/gcc- lib/mingw32/3.3.3/specs Configured with: ../gcc/configure --with-gcc --with-gnu- ld --with-gnu-as --host=mingw32 --target=mingw32 -- prefix=/mingw --enable-threads --disable-nls --enable- languages=c,c++,f77,ada,objc,java --disable-win32- registry --disable-shared --enable-sjlj-exceptions -- enable-libgcj --disable-java-awt --without-x --enable- java-gc=boehm --disable-libgcj-debug --enable- interpreter --enable-hash-synchronization Thread model: win32 gcc version 3.3.3 (mingw special) e@TRUFFLE ~ $ ld -v GNU ld version 2.15.90 20040222 e@TRUFFLE ~ $ uname -a MINGW32_NT-5.1 TRUFFLE 1.0.10(0.46/3/2) 2004-03- 010 08:07 i686 unknown installed from gcc-core-3.3.3-20040217-1.tar.gz gcc-g++-3.3.3-20040217-1.tar.gz binutils-2.15.90-20040222-1.tar.gz MSYS-1.0.10.exe e ---------------------------------------------------------------------- >Comment By: Danny Smith (dannysmith) Date: 2004-04-25 17:31 Message: Logged In: YES user_id=11494 Confirmed. Bug not present on gcc 3.4 or trunk. The problem is with stack probe (aka __alloca in libgcc.a). A patch was backported from trunk to fix a problem with regparm attribute vs __alloca. I'll test to see if reverting that patch fixes the -O3 bug Danny ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=102435&aid=940940&group_id=2435 |
From: SourceForge.net <no...@so...> - 2004-04-25 08:39:56
|
Bugs item #940940, was opened at 2004-04-24 07:45 Message generated for change (Comment added) made by dannysmith You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=102435&aid=940940&group_id=2435 Category: gcc Group: Known bugs Status: Open Resolution: None >Priority: 9 Submitted By: Doug Currie (dougcurrie) Assigned to: Danny Smith (dannysmith) Summary: optimization error building zlib1.2.1 Initial Comment: With gcc.exe (GCC) 3.3.3 (mingw special) compiling the zlib 1.2.1 distribution from http://www.gzip.org/zlib/ with the zlib supplied win32/Makefile.gcc the built-in tests fail. The problem occurs in the function int ZEXPORT gzclose (file) wherein the value read for s->mode is incorrect. If in win32/Makefile.gcc I replace CFLAGS = $(LOC) -O3 -Wall with CFLAGS = $(LOC) -O2 -Wall the tests pass. I conclude that gcc.exe (GCC) 3.3.3 (mingw special) has an optimization related bug. Other configuration info: Windows XP sp1 e@TRUFFLE ~ $ gcc -v Reading specs from c:/msys/mingw/bin/../lib/gcc- lib/mingw32/3.3.3/specs Configured with: ../gcc/configure --with-gcc --with-gnu- ld --with-gnu-as --host=mingw32 --target=mingw32 -- prefix=/mingw --enable-threads --disable-nls --enable- languages=c,c++,f77,ada,objc,java --disable-win32- registry --disable-shared --enable-sjlj-exceptions -- enable-libgcj --disable-java-awt --without-x --enable- java-gc=boehm --disable-libgcj-debug --enable- interpreter --enable-hash-synchronization Thread model: win32 gcc version 3.3.3 (mingw special) e@TRUFFLE ~ $ ld -v GNU ld version 2.15.90 20040222 e@TRUFFLE ~ $ uname -a MINGW32_NT-5.1 TRUFFLE 1.0.10(0.46/3/2) 2004-03- 010 08:07 i686 unknown installed from gcc-core-3.3.3-20040217-1.tar.gz gcc-g++-3.3.3-20040217-1.tar.gz binutils-2.15.90-20040222-1.tar.gz MSYS-1.0.10.exe e ---------------------------------------------------------------------- >Comment By: Danny Smith (dannysmith) Date: 2004-04-25 20:39 Message: Logged In: YES user_id=11494 Yup, it is an _alloca bug, and is fixed in 3.3.3 by reverting this patch: 2003-11-20 Christopher Faylor <cg...@re...> Backport from trunk 2003-10-25 Richard Henderson <rt...@re...> * config/i386/i386.c (ix86_eax_live_at_start_p): New. (ix86_expand_prologue): Save and restore eax around stack probe if it's live. Actually, just adding the emit_insn (gen_blockage (const0_rtx)); line back in fixes this bug IMO, this is a show-stopper for pushing 3.3.3 into current. Danny ---------------------------------------------------------------------- Comment By: Danny Smith (dannysmith) Date: 2004-04-25 17:31 Message: Logged In: YES user_id=11494 Confirmed. Bug not present on gcc 3.4 or trunk. The problem is with stack probe (aka __alloca in libgcc.a). A patch was backported from trunk to fix a problem with regparm attribute vs __alloca. I'll test to see if reverting that patch fixes the -O3 bug Danny ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=102435&aid=940940&group_id=2435 |
From: SourceForge.net <no...@so...> - 2004-07-26 11:01:16
|
Bugs item #940940, was opened at 2004-04-24 07:45 Message generated for change (Settings changed) made by dannysmith You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=102435&aid=940940&group_id=2435 Category: gcc Group: Known bugs >Status: Closed >Resolution: Fixed Priority: 9 Submitted By: Doug Currie (dougcurrie) Assigned to: Danny Smith (dannysmith) Summary: optimization error building zlib1.2.1 Initial Comment: With gcc.exe (GCC) 3.3.3 (mingw special) compiling the zlib 1.2.1 distribution from http://www.gzip.org/zlib/ with the zlib supplied win32/Makefile.gcc the built-in tests fail. The problem occurs in the function int ZEXPORT gzclose (file) wherein the value read for s->mode is incorrect. If in win32/Makefile.gcc I replace CFLAGS = $(LOC) -O3 -Wall with CFLAGS = $(LOC) -O2 -Wall the tests pass. I conclude that gcc.exe (GCC) 3.3.3 (mingw special) has an optimization related bug. Other configuration info: Windows XP sp1 e@TRUFFLE ~ $ gcc -v Reading specs from c:/msys/mingw/bin/../lib/gcc- lib/mingw32/3.3.3/specs Configured with: ../gcc/configure --with-gcc --with-gnu- ld --with-gnu-as --host=mingw32 --target=mingw32 -- prefix=/mingw --enable-threads --disable-nls --enable- languages=c,c++,f77,ada,objc,java --disable-win32- registry --disable-shared --enable-sjlj-exceptions -- enable-libgcj --disable-java-awt --without-x --enable- java-gc=boehm --disable-libgcj-debug --enable- interpreter --enable-hash-synchronization Thread model: win32 gcc version 3.3.3 (mingw special) e@TRUFFLE ~ $ ld -v GNU ld version 2.15.90 20040222 e@TRUFFLE ~ $ uname -a MINGW32_NT-5.1 TRUFFLE 1.0.10(0.46/3/2) 2004-03- 010 08:07 i686 unknown installed from gcc-core-3.3.3-20040217-1.tar.gz gcc-g++-3.3.3-20040217-1.tar.gz binutils-2.15.90-20040222-1.tar.gz MSYS-1.0.10.exe e ---------------------------------------------------------------------- Comment By: Danny Smith (dannysmith) Date: 2004-04-25 20:39 Message: Logged In: YES user_id=11494 Yup, it is an _alloca bug, and is fixed in 3.3.3 by reverting this patch: 2003-11-20 Christopher Faylor <cg...@re...> Backport from trunk 2003-10-25 Richard Henderson <rt...@re...> * config/i386/i386.c (ix86_eax_live_at_start_p): New. (ix86_expand_prologue): Save and restore eax around stack probe if it's live. Actually, just adding the emit_insn (gen_blockage (const0_rtx)); line back in fixes this bug IMO, this is a show-stopper for pushing 3.3.3 into current. Danny ---------------------------------------------------------------------- Comment By: Danny Smith (dannysmith) Date: 2004-04-25 17:31 Message: Logged In: YES user_id=11494 Confirmed. Bug not present on gcc 3.4 or trunk. The problem is with stack probe (aka __alloca in libgcc.a). A patch was backported from trunk to fix a problem with regparm attribute vs __alloca. I'll test to see if reverting that patch fixes the -O3 bug Danny ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=102435&aid=940940&group_id=2435 |