=?utf-8?Q?Re=3A_[Gc]_Test_failures, _hangs, _crashes_with_2011-04-25_CVS_checkout?=

Ivan Maidanski ivmai at mail.ru
Tue Apr 26 14:18:09 PDT 2011


Hi Peter,

Mon, 25 Apr 2011 21:10:40 -0500 "Peter O'Gorman" <gc at mlists.thewrittenword.com>:

> Hi,
> 
> Building from today's cvs in all cases with --enable-threads=posix, and
> with -DAO_USE_PTHREAD_DEFS on AIX, IRIX, Tru64 and Solaris 10/x86 (older
> sun studio can't build the inline asm).
> 
> On AIX 5.1 every test fails like this:
> FAIL: hugetest
> sem_init failed
> /bin/sh: 1335468 IOT/Abort trap(coredump)

Fixed (I hope). Try -DGC_SEM_INIT_PSHARED=1
If it works ok then I'll add -D to configure for AIX pre-5.3.

> 
> AIX 5.3, 6.1 and 7.1 pass all tests.
> 
> initsecondarythread crashes on all HP-UX and on Tru64.
> On ia64-hp-hpux11.31 the crash is here:
> SEGV_MAPERR - Address not mapped to object
> #0  0x60000000ca58ae20:0 in GC_mark_from (mark_stack_top=0x40058040, 
> mark_stack=0x40058000, mark_stack_limit=0x40060000) at mark.c:820
> 820               deferred = *(word *)limit;
> (gdb) p limit
> $2 = 0x777b50b8 <Address 0x777b50b8 out of bounds>

This happens due to GC_get_main_stack_base uses hard-coded STACKBOTTOM (which is for the main thread). So, GC_INIT should be called from the primordial thread only on these platforms.
Test fixed. Please verify.

> 
> Solaris 7 and 8 , IRIX 6.5, AIX 5.2, and sometimes on HP-UX gctest hangs
> forever.

More details please.

> 
> Also, strangely, when building with --disable-threads, Solaris 10/x86
> gctest crashes with a segv, every other one of our platforms passes all
> tests. segv is in test.c at
> if (SEXPR_TO_INT(car(car(list))) != low) {

More details please: stack trace, value of low, value of car(car(list)), call print_int_list.

Regards.

> 
> Let me know if you need more info on any of these.
> 
> Thanks,
> Peter
> -- 
> Peter O'Gorman
> pogma at thewrittenword.com



More information about the Gc mailing list