[kaffe] m68k/linux with jthreads

Kiyo Inaba inaba at src.ricoh.co.jp
Sun Jan 16 16:29:07 PST 2005


Hi,

I was frustrated for having some non-deterministic errors happened on
this platform for long time. I finally can figure out why it happens
(in some sence). It was caused by wrong behavior of 'setitimer' for
m68k/linux. So my proposed patch attacheched disable the use of
'setitimer' for this platform. This patch solves above error with
jthreads and the regression tests 'only' fails 46 test cases.
# BTW, recently pthreads works for this platform with 44 fails.

ChangeLog entry can be:
------------------------------------------------------------------
2005-01-17	Kiyo Inaba	<inaba at src.ricoh.co.jp>

	* config/m68k/linux/config.frag:
	Disable the use of 'setitimer' even if it is available.
------------------------------------------------------------------

Kiyo

diff -Naur kaffe-snap-041209/config/m68k/linux/config.frag ../kaffe-snap-041209/config/m68k/linux/config.frag
--- kaffe-snap-041209/config/m68k/linux/config.frag	2003-08-04 18:20:04.000000000 +0900
+++ ../kaffe-snap-041209/config/m68k/linux/config.frag	2005-01-12 13:20:41.000000000 +0900
@@ -3,6 +3,10 @@
 #
 CFLAGS="$CFLAGS -fno-omit-frame-pointer"
 
+# I know setitimer is available on this platform, but it is not compatible
+# (carefully I omit to say 'setitimer is buggy') with kaffe's jthread.
+ac_cv_func_setitimer=${ac_cv_func_setitimer='no'}
+
 # if we use cross environment, following values may not be detected.
 if [ "$cross_compiling" = yes ]; then
   ac_cv_alignmentof_voidp=${ac_cv_alignmentof_voidp='2'}







More information about the kaffe mailing list