Bug 42981

Summary:	internal compiler error: in default_secondary_reload, at targhooks.c:660
Product:	gcc	Reporter:	Khem Raj <raj.khem>
Component:	target	Assignee:	Not yet assigned to anyone <unassigned>
Status:	NEW ---
Severity:	normal	CC:	gcc-bugs, mikpelinux, ramana
Priority:	P3	Keywords:	ice-on-valid-code
Version:	4.4.4
Target Milestone:	---
Host:	x86_64-linux-gnu	Target:	arm-none-eabi
Build:	x86_64-linux-gnu	Known to work:	4.5.0
Known to fail:	4.4.2 4.4.4	Last reconfirmed:	2010-02-06 13:21:34
Attachments:	testcase Reduced test case

Description Khem Raj 2010-02-06 00:05:08 UTC

gcc ICE's when compiling attached example

arm-none-eabi-gcc lib_pad.i -c -O  -fPIC -mthumb

../ncurses/./base/lib_pad.c: In function ‘pnoutrefresh’:
../ncurses/./base/lib_pad.c:218: internal compiler error: in default_secondary_reload, at targhooks.c:660
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

Target: arm-none-eabi
Configured with: ../sources/gcc-4_4-branch/configure --target=arm-none-eabi --prefix=/scratch/oss/baremetal/arm-none-eabi/tools --enable-languages=c,c++ --with-newlib
Thread model: single
gcc version 4.4.4 20100203 (prerelease) (GCC)

Comment 1 Khem Raj 2010-02-06 00:06:26 UTC

Created attachment 19811 [details]
testcase

Comment 2 Richard Earnshaw 2010-02-06 13:21:34 UTC

Confirmed

Comment 3 Richard Earnshaw 2010-02-06 13:22:25 UTC

Created attachment 19813 [details]
Reduced test case

Comment 4 Mikael Pettersson 2010-02-06 21:08:00 UTC

A bisection has identified this as a regression on 4.4 branch triggered by r147081, the fix for PR39914. The same change was committed to trunk too, but for some reason doesn't trigger the ICE there.

Comment 5 Mikael Pettersson 2010-02-07 11:17:23 UTC

(In reply to comment #4)
> A bisection has identified this as a regression on 4.4 branch triggered by
> r147081, the fix for PR39914. The same change was committed to trunk too, but
> for some reason doesn't trigger the ICE there.

Trunk also ICEd on this test case for a while, but that was fixed by r147282,
one of several fixes to fwprop/df-problems.

Comment 6 Mikael Pettersson 2010-02-07 11:54:19 UTC

gcc-4.5-20100204 ICEs with -O1 -fno-forward-propagate -fPIC -mthumb on the original test case. So r147282, which enabled fwprop at -O1, merely masked the problem.