ICE in arm_select_dominance_cc_mode, at config/arm/arm.c:10625

[Matthias Klose]

seen with 4.5 and 4.6 with -O1. Other optimization levels don't show the ICE.

$ gfortran -c -O1 -fpermissive -Wfatal-errors -w fbasics_gss-min.f
fbasics_gss-min.f: In function 'deval':
fbasics_gss-min.f:39:0: internal compiler error: in arm_select_dominance_cc_mode, at config/arm/arm.c:11474
Please submit a full bug report,
with preprocessed source if appropriate.

trunk 20120107 has a different ICE:
$ /usr/lib/gcc-snapshot/bin/gfortran -c -O1 -fpermissive -Wfatal-errors -w fbasics_gss-min.f
fbasics_gss-min.f:139:0: internal compiler error: vector VEC(tree,base) index domain error, in ipa_get_indirect_edge_target at ipa-cp.c:1115
Please submit a full bug report,
with preprocessed source if appropriate.

[Doko-v (doko-v)]

Created attachment 26784
test case

seen with 4.5 and 4.6 with -O1. Other optimization levels don't show the ICE.

$ gfortran -c -O1 -fpermissive -Wfatal-errors -w fbasics_gss-min.f
fbasics_gss-min.f: In function 'deval':
fbasics_gss-min.f:39:0: internal compiler error: in arm_select_dominance_cc_mode, at config/arm/arm.c:11474
Please submit a full bug report,
with preprocessed source if appropriate.

trunk 20120107 has a different ICE:
$ /usr/lib/gcc-snapshot/bin/gfortran -c -O1 -fpermissive -Wfatal-errors -w fbasics_gss-min.f
fbasics_gss-min.f:139:0: internal compiler error: vector VEC(tree,base) index domain error, in ipa_get_indirect_edge_target at ipa-cp.c:1115
Please submit a full bug report,
with preprocessed source if appropriate.

[Asa Sandahl (asa-sandahl)]

Thank you for the bug report. I've confirmed this with gcc-linaro-4.6-2012.02 on ARM:

/home/asa-san/bugs/cbuild/slaves/ursa4/gcc-linaro-4.6-2012.02 gfortran -c -O1 -fpermissive -Wfatal-errors -w fbasics_gss-min.f
fbasics_gss-min.f: In function 'deval':
fbasics_gss-min.f:39:0: internal compiler error: in arm_select_dominance_cc_mode, at config/arm/arm.c:11452
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions

The work-around is to compile at -O2.

The fault exists in all versions of gcc-linaro-4.6 and gcc-linaro-4.5 that I have tried, from 2011.04 and forward on.
The fault also exists in gcc-4.5.3 and gcc-4.6.2.

On trunk there is another error message, as stated in the ticket. However, this error does not go away when switching to -O2 or -03. The work around in this case is to compile at -O0.
Running the test over a few older versions of 4.7 shows that the behavior changed somewhere along the way.

Old error message still shows on 177688:
/tools/toolchains/arch/armv7l/gcc-4.7~svn177688-armv7l-natty-cbuild161-ursa1-cortexa9r1 gfortran -c -O1 -fpermissive -Wfatal-errors -w fbasics_gss-min.f
fbasics_gss-min.f: In function 'deval':
fbasics_gss-min.f:39:0: internal compiler error: in arm_select_dominance_cc_mode, at config/arm/arm.c:10902
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

The new message is seen on 182152:
/tools/toolchains/arch/armv7l/gcc-4.7~svn182152-armv7l-natty-cbuild212-tcpanda02-cortexa9r1 gfortran -c -O1 -fpermissive -Wfatal-errors -w fbasics_gss-min.f
fbasics_gss-min.f:139:0: internal compiler error: vector VEC(tree,base) index domain error, in ipa_get_indirect_edge_target at ipa-cp.c:1115
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

The same error occurs at -O2 (and -O3).
/tools/toolchains/arch/armv7l/gcc-4.7~svn182152-armv7l-natty-cbuild212-tcpanda02-cortexa9r1 gfortran -c -O2 -fpermissive -Wfatal-errors -w fbasics_gss-min.f
fbasics_gss-min.f:139:0: internal compiler error: vector VEC(tree,base) index domain error, in ipa_get_indirect_edge_target at ipa-cp.c:1115
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

Further investigation will show if this is question of two different errors.

I've set this ticket to medium priority as it is a ftbfs, there is a work-around and the error probably exists upstream.

[Michael Collison (michael-collison)]

Will not fix. Does not occur with linaro 4.8 or 4.9.

[Ktkachov (ktkachov)]

I can't reproduce this with 4.8.3, 4.9.1 or 4.10 trunk.

Is this still an issue?

[Doko-v (doko-v)]

yes, can't reproduce anymore with 4.9 and trunk