http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52074
Jakub Jelinek <jakub at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jakub at gcc dot gnu.org --- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-02-01 11:47:00 UTC --- C testcase (-m64 -O2): struct S { const char *d, *e; } __attribute__((packed)); void foo (const char **p, struct S *q) { *p = "abcdef"; q->d = "ghijk"; } I'd say the bug is in the section anchor support (and the bug goes away with -O2 -fno-section-anchors; so I'd expect this is a regression since -fsection-anchors has been introduced and made default on ppc64), because expand_normal on ADDR_EXPR of the STRING_CST returns a PLUS (which IMHO should be ok only for EXPAND_SUM etc. but not EXPAND_NORMAL).