[kaffe] CVS kaffe (guilhem): GC Deadlock fixes for Linux 2.4/2.6 (at least).

[Davanum Srinivas]

if someone gives me a patch that uses glibc's backtrace or libunwind to print more information
during assert...i'd be happy to apply the local patch and try to recreate the problem on brutus.

thanks,
dims

--- Davanum Srinivas <dims at yahoo.com> wrote:

> usual problem as before. I can confirm that regression tests works fine on brutus.
> 
> *** Semantic Warning: This switch block can fall through to the next case. Did you forget a
> break
> statement?
> ... Copying Required Files
> ... Building Ant Distribution
> Buildfile: build.xml
> kaffe-bin: utf8const.c:58: do_lockUTF: Assertion `utfLockRoot == ((void *)0)' failed.
> /usr/local/gump/kaffe/workspace/ant/bootstrap.sh: line 153: 27398 Aborted                
> "${JAVACMD}" -classpath "${CLASSPATH}" -Dant.home=. $ANT_OPTS org.apache.tools.ant.Main -emacs
> "$@" bootstrap
> ... Failed Building Ant Distribution !
> 
> -- dims
> 
> --- Adam Heath <doogie at brainfood.com> wrote:
> 
> > On Sun, 2 Jan 2005, Kaffe CVS wrote:
> > 
> > > PatchSet 5747
> > > Date: 2005/01/02 19:03:30
> > > Author: guilhem
> > > Branch: HEAD
> > > Tag: (none)
> > > Log:
> > > GC Deadlock fixes for Linux 2.4/2.6 (at least).
> > >
> > >         * kaffe/kaffevm/systems/unix-pthreads/lock-impl.c
> > >         (setBlockState): Protect the mutex locking against signal.
> > >
> > >         * kaffe/kaffevm/systems/unix-pthreads/thread-impl.c,
> > >         kaffe/kaffevm/systems/unix-pthreads/thread-internal.h
> > >         (KaffePThread_AckAndWaitForResume): New function.
> > >         (suspend_handler): Call directly KaffePThread_AckAndWaitForResume
> > >         now.
> > >         (KaffePThread_getSuspendSignal): Export the signal number
> > >         corresponding to the suspend signal.
> > >         (jthread_suspendall): Cleant up the semaphore sweep. Added a few
> > >         assertion checking.
> > >         (jthread_unsuspendall): Added a few assertion checking.
> > >         (tRun): Check whether the thread is killed when it is woken up.
> > >         (jthread_exit): Do not cancel any thread as it may cause deadlocks
> > >         on 2.4. Only interrupt and set the state to THREAD_KILL.
> > >
> > >         * libraries/clib/native/Thread.c
> > >         (VMThread.sleep): Check that if timeout is null we should return
> > >         immediately.
> > 
> > This change has appeared to make all test cases pass on my smp system.  Well
> > done.
> > 
> > However, Dalibor has reported an assertion failure for bootstrap ant on gump.
> > 
> > _______________________________________________
> > kaffe mailing list
> > kaffe at kaffe.org
> > http://kaffe.org/cgi-bin/mailman/listinfo/kaffe
> > 
> 
> 
> =====
> Davanum Srinivas - http://webservices.apache.org/~dims/
> 
> 
> 		
> __________________________________ 
> Do you Yahoo!? 
> Yahoo! Mail - Find what you need with new enhanced search.
> http://info.mail.yahoo.com/mail_250
> 
> _______________________________________________
> kaffe mailing list
> kaffe at kaffe.org
> http://kaffe.org/cgi-bin/mailman/listinfo/kaffe
> 


=====
Davanum Srinivas - http://webservices.apache.org/~dims/


		
__________________________________ 
Do you Yahoo!? 
Yahoo! Mail - 250MB free storage. Do more. Manage less. 
http://info.mail.yahoo.com/mail_250