https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126518
Bug ID: 126518
Summary: Execution FAIL of gcc.dg/constructor-1.c on cris-elf
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: roger at nextmovesoftware dot com
Target Milestone: ---
The testcase gcc.dg/constructor-1.c (and related global constructor tests)
currently FAILs at run-time on cris-elf (and can be observed in a cross-build
with RUNTESTFLAGS="--target_board=cris-sim"). This may be a regression caused
by (or related to) the fix for PR rtl-optimization/126276. However, my
attempts to bisect the issue appear to show that we produce idential assembly
language for this testcase as we did in gcc-16.1.0. I'm filing this as a PR in
bugzilla, so the issue can be tracked, including determining whether this is
indeed a regression (or just a known failure). The related
c-c++-common/init-pri1.c (and 5 other init-pri1*.c testcases) are also failing
are runtime.
constructor-1.c doesn't use the *btst instruction, so is unlikely to be
affected by PR126276. This might be a change/breakage in newlib's
crtbegin.o/crtend.o [but I'm out of my depth].
