[Bug tree-optimization/41031] [4.5 Regression]: build breakage for cris-elf building newlib, ICE in insert_value_copy_on_edge

2009-08-11 Thread krebbel at gcc dot gnu dot org
--- Comment #3 from krebbel at gcc dot gnu dot org 2009-08-11 07:56 --- The same happens when bootstrapping on s390x while building libdecnumber. Here is a reduced testcase: typedef struct { unsigned char bits; unsigned short int lsu[1]; } decNumber; void decCompareOp (decNumber *

[Bug tree-optimization/41031] [4.5 Regression]: build breakage for cris-elf building newlib, ICE in insert_value_copy_on_edge

2009-08-11 Thread hp at gcc dot gnu dot org
--- Comment #4 from hp at gcc dot gnu dot org 2009-08-11 09:26 --- (In reply to comment #3) The same happens when bootstrapping on s390x while building libdecnumber. Here is a reduced testcase: At -O1, the simplified test-case exposes the bug for cris-elf too. FWIW, I guess you must

[Bug tree-optimization/41031] [4.5 Regression]: build breakage for cris-elf building newlib, ICE in insert_value_copy_on_edge

2009-08-10 Thread hp at gcc dot gnu dot org
--- Comment #1 from hp at gcc dot gnu dot org 2009-08-11 00:18 --- Created an attachment (id=18339) -- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18339action=view) Preprocessed ecvtbuf.c from months-old newlib Compile with -O2 to reproduce ICE. --

[Bug tree-optimization/41031] [4.5 Regression]: build breakage for cris-elf building newlib, ICE in insert_value_copy_on_edge

2009-08-10 Thread hp at gcc dot gnu dot org
--- Comment #2 from hp at gcc dot gnu dot org 2009-08-11 01:02 --- I had a brief look. Seems like the new code assumes that promote_mode always promotes the input mode (ahem! :) when mismatching, which it only does if the target defines PROMOTE_MODE (which CRIS does not; it just has a