Bug 224

Summary:	Malloc_Test seg faults on LynxOS
Product:	ACE	Reporter:	levine
Component:	ACE Tests	Assignee:	Douglas C. Schmidt <schmidt>
Status:	RESOLVED ---
Severity:	normal
Priority:	P2
Version:	5.0
Hardware:	All
OS:	All

Description levine 1999-08-10 10:25:46 CDT

Malloc_Test seg faults here on LynxOS, both x86 and PPC:

#0  0x1f97 in ACE_Malloc<ACE_MMAP_Memory_Pool, ACE_MMAP_Memory_Pool_Options,
ACE_Process_Mutex>::shared_malloc ()
#1  0x2162 in ACE_Malloc<ACE_MMAP_Memory_Pool, ACE_MMAP_Memory_Pool_Options,
ACE_Process_Mutex>::malloc ()
#2  0x4b1 in initialize ()

Comment 1 levine 1999-08-10 14:23:59 CDT

The SEGV happens:

    at /project/danzontmp/levine/ACE_wrappers/build/lynx/ace/Malloc_T.cpp:316
316           if (currp->s_.size_ >= nunits) // Big enough

currp appears to have a bogus value:

(gdb) p currp
$4 = (ACE_Malloc_Header *) 0xf68990c3
(gdb) p *currp
Cannot access memory at address 0xf68990c3.

Comment 2 levine 1999-08-11 11:52:59 CDT

From Doug:
        That's weird!  The only thing I can think of is that maybe
LynxOS doesn't like this address in the child:

static const void *CHILD_BASE_ADDR =
#if defined (ACE_HAS_POSITION_INDEPENDENT_MALLOC)
       64 * 1024 +
#endif /* ACE_HAS_POSITION_INDEPENDENT_MALLOC */
       ACE_DEFAULT_BASE_ADDR;

Can you please let me know if the segfault is occurring in the child
or in the parent?  If it's in the parent, then I'm at a loss.  If it's
in the child, then we can probably pick a better base address, which
should make LynxOS happy.

Comment 3 levine 1999-08-11 12:00:59 CDT

The segfault is in the parent.

ACE_DEFAULT_BASE_ADDR is 0 on LynxOS.  That tells the
OS to pick the address.  It works well best other tests.
ACE_MAP_FIXED is 0, as well.

Comment 4 Douglas C. Schmidt 1999-08-27 18:30:59 CDT

I recommend that we don't enable ACE_HAS_POSITION_INDEPENDENT_MALLOC with
Lynx.

Comment 5 levine 1999-08-27 19:31:59 CDT

As far as I can tell, we don't enable ACE_HAS_POSITION_INDEPENDENT_MALLOC
on LynxOS.

Comment 6 Douglas C. Schmidt 1999-10-06 08:43:59 CDT

It's not clear why this fails, so for now, I've commented out the test on
LynxOS

Comment 7 levine 1999-10-13 10:15:59 CDT

Fixed the disabling of the test on LynxOS:
ChangeLogTag: Wed Oct 13 10:11:47 1999  David L. Levine  <levine@cs.wustl.edu>