Hi,

I think that adding synchronization in module.c patch is just fine here. There is nothing global being used in the code as far as I can tell that would prevent it from being re-entrant. So the calls to NSAddImage which are internal MacOSX routines must be the culprit here. It would be interesting to look at the libdl code to see if it uses a mutex when using similar functions. Any additional overhead of obtaining a mutex would be meaningless next to the time to actually load any shared library.

I would go ahead and commit the change since it can do no damage.

I will try to track down the libdl source to see if it does something similar. I don't think the code that implements NSAddImage is available to be examined. Perhaps it is missing the required "sync" "isync" pair needed when loading code into what was previously in the data cache (requires a good cache flush as well) that a mutex implementation would provide.

Does anyone know is the source code to NSAddImage and other related functions are available via the Darwin project?

Kevin




On Oct 21, 2005, at 2:23 AM, Tino Rachui - Sun Germany - Development - Software Engineer wrote:

Hi Kevin,

please have a look at http://porting.openoffice.org/servlets/ ReadMsg?list=dev&msgId=1961840
(see also #i55863#).

Kind regards,
Tino

Kevin B. Hendricks wrote:

Hi,
I have been using an m133 based build on a dual processor G5 machine with no problems so far. My usage has not been extensive. I assume this must be a cache or synchronization/race based issue. Can you simply post the patch so that I can examine it to see what it does and why?
Thanks,
Kevin
On Oct 18, 2005, at 3:03 PM, Tino Rachui wrote:

Hi!

Are there (lucky) owners of multi-processor machines out there who have installed the one of the latest versions of OOo2.0 for Mac OS X (>=SRC680m130)?
Patrick Luby the man behind NeoOffice/J reported that there exist
serious problems during loading of dynamic libraries on multi- processor machines (see #i55863#). He kindly donated a patch to fix these problems. However before introducing this patch into the OOo2.0 code line I would like to know if the problems still exist. So anyone who has a multi-processor: Did you noticed sporadic crashes or deadlocks when using OOo2.0?
If yes would you please contact me?
Thanks in advance.

Kind regards,
Tino

-------------------------------------------------------------------- -
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to