g++ ICE: in elimination_costs_in_insn, at reload1.c:3633

[Xerxes Rånby]

Use the attached preprocessed compile.i source code to trigger the bug.

g++ compile.i
In file included from src/machine.h:14:0,
                 from src/compile.cpp:11:
src/common.h:279:1: note: the mangling of ‘va_list’ has changed in GCC 4.4
src/compile.cpp: In function ‘void {anonymous}::local::compile({anonymous}::local::MyThread*, {anonymous}::local::Frame*, unsigned int, int)’:
src/compile.cpp:6211:1: internal compiler error: in elimination_costs_in_insn, at reload1.c:3633
Please submit a full bug report,
with preprocessed source if appropriate.
See <file:///usr/share/doc/gcc-4.6/README.Bugs> for instructions.
Preprocessed source stored into /tmp/ccPPh5os.out file, please attach this to your bugreport.

#The preprocessed compile.i got generated during a debug build of avian
#Full avian debug build instructions below:
apt-get openjdk-7-jdk libz-dev git
git clone https://github.com/ReadyTalk/avian.git
cd avian
JAVA_HOME=/usr/lib/jvm/java-7-openjdk-armhf/ make openjdk=/usr/lib/jvm/java-7-openjdk-armhf/ mode=debug

[Belagod (belagod)]

The attached test causes gcc to ICE when compiled with -O0.

$ arm-eabi-gcc -c foo.c
foo.c: In function 'f0a':
foo.c:16:1: warning: left shift count >= width of type [enabled by default]
foo.c:16:1: warning: left shift count >= width of type [enabled by default]
foo.c:16:1: warning: right shift count >= width of type [enabled by default]
foo.c:16:1: warning: right shift count >= width of type [enabled by default]
foo.c:16:1: warning: right shift count >= width of type [enabled by default]
foo.c:16:1: warning: left shift count >= width of type [enabled by default]
foo.c:16:1: warning: left shift count >= width of type [enabled by default]
foo.c:16:1: warning: right shift count >= width of type [enabled by default]
foo.c:16:1: warning: right shift count >= width of type [enabled by default]
foo.c:16:1: warning: right shift count >= width of type [enabled by default]
foo.c:18:1: internal compiler error: in elimination_costs_in_insn, at reload1.c:3643
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

[Belagod (belagod)]

Created attachment 24002
test case with complex expressions

[Rguenth (rguenth)]

Please specify the compiler version you used.

[Mikpe (mikpe)]

I can trigger the ICE on armv5tel-unknown-linux-gnueabi with gcc-4.7-20110416 and gcc-4.6-20110415 if I pass both -O0 and -march=armv7-a. Doesn't ICE with -O2 or -march=armv5te. Also doesn't ICE with gcc-4.5 or gcc-4.4 regardless of options.

[Belagod (belagod)]

Sorry for the delay: this is my gcc -v:

arm-eabi-gcc
COLLECT_LTO_WRAPPER=.../trunk/2011-04-20/install-a9softfpneon/bin/../libexec/gcc/arm-eabi/4.7.0/lto-wrapper
Target: arm-eabi
Configured with: .../trunk/source/gcc-fsf/configure --target=arm-eabi --enable-checking=release --disable-gdbtk --disable-werror --disable-tui --disable-rda --disable-sid --disable-utils --disable-lto --with-cpu=cortex-a9 --with-float=softfp --with-fpu=neon --prefix=.../trunk/2011-04-20/install-a9softfpneon --with-mpfr=.../tools/linux_x86_64 --with-mpc=.../tools/linux_x86_64 --with-gmp=.../tools/linux_x86_64 --disable-lto --disable-werror --disable-shared --disable-nls --disable-threads --disable-tls --enable-checking=yes --enable-languages=c,c++,fortran --with-newlib
Thread model: single
gcc version 4.7.0 20110420 (experimental) (GCC)

[Xerxes Rånby (xranby)]

This bug is reproducible using:
g++ --version
g++ (Debian 4.6.3-8+rpi1) 4.6.3

[Xerxes Rånby (xranby)]

pi@raspberrypi ~ $ g++ -v
Using built-in specs.
COLLECT_GCC=g++
COLLECT_LTO_WRAPPER=/usr/lib/gcc/arm-linux-gnueabihf/4.6/lto-wrapper
Target: arm-linux-gnueabihf
Configured with: ../src/configure -v --with-pkgversion='Debian 4.6.3-8+rpi1' --with-bugurl=file:///usr/share/doc/gcc-4.6/README.Bugs --enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.6 --enable-shared --enable-linker-build-id --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.6 --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-gnu-unique-object --enable-plugin --enable-objc-gc --disable-sjlj-exceptions --with-arch=armv6 --with-fpu=vfp --with-float=hard --enable-checking=release --build=arm-linux-gnueabihf --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf
Thread model: posix
gcc version 4.6.3 (Debian 4.6.3-8+rpi1)

[Xerxes Rånby (xranby)]

The crash looks remotely similar to the upsteam gcc bug

[peter green (plugwash)]

"The crash looks remotely similar to the upsteam gcc bug"
I don't think the two bugs are the same. The upstream bug you linked is reported as happening in 4.7 while this bug only happens with 4.6.

I do not intend to fix or push upstream bugs in 4.6 that do not also happen in 4.7. Please just use 4.7 in such cases.