[Gc] FW from Rainer Orth: Further investigation of Solaris 10/x86 gctest failure

Boehm, Hans hans.boehm at hp.com
Thu Jun 9 12:13:22 PDT 2005


The moderator ( :-) ) mistakenly rejected this the first time.

Hans

-----Original Message-----
From: Rainer Orth [mailto:ro at TechFak.Uni-Bielefeld.DE] 
Sent: Monday, June 06, 2005 3:13 PM
To: Boehm, Hans
Cc: gc at napali.hpl.hp.com
Subject: Further investigation of Solaris 10/x86 gctest failure


I've just posted a GCC bug report about my findings concerning the
gctest failure observed on Solaris 10/x86:

	http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21940

I'm not sure how to best proceed from here (and where to discuss the
further investigation, either on gcc-bugs/GCC bugzilla, the gc list or
in private mail).

Anyway, I think there are a couple of important points:

* It seems like the integration of amd64 support in Solaris 10 broke
  boehm-gc.

* Without calls to GC_enable_incremental() (or with --enable-full-debug)
and
  without thread support, gctest works.

* Without thread support, gctest fails like this (both with cc and gcc):

Switched to incremental mode
Reading dirty bits from /proc
List reversal produced incorrect list - collector is broken Test failed
Abort

* With thread support, the failure scenario is different (but again
  identical with cc and gcc):

Apparently failed to mark form some function arguments.
Perhaps GC_push_regs was configured incorrectly?
Test failed
Abort

* Defining USE_GENERIC_PUSH_REGS doesn't make a difference.

	Rainer

------------------------------------------------------------------------
-----
Rainer Orth, Faculty of Technology, Bielefeld University



More information about the Gc mailing list