On Dec 15, 2006, at 19:52, Sean Bryant wrote:
> Sean Bryant wrote:
>> SASADA Koichi wrote:
>>> Hi,
>>>
>>> Sean Bryant wrote:
>>>> Hi I decided i'd try and get the some snapshot of yarv up and  
>>>> running.
>>>> It compiles with out complaining but the test fail and generator  
>>>> a core.
>>>>  The core says the problem is in signalcontext() in libc. But I  
>>>> doubt
>>>> that's the case. That's just the last running thread. Switching to
>>>> thread 3, the main thread it seems reveals nothing more than the  
>>>> thread
>>>> sleeping just as gdb reports. Any tips on debugging this?
>>>
>>> What's test?  "make test"?
>>>
>> Yes, it is 'make test'. I'm running FreeBSD-6.2-RC1 under QEMU. I  
>> tried the same under a 6.1-STABLE install and the same thing  
>> happened.
>
> To update the list. Forcing a link against libthr by exporting  
> LDFLAGS='-lthr' will no longer have yarv core on a fork.

No, no, no, no, no.

Use libmap.conf.  This is exactly what it exists for.

> It does however introduce new and interesting problems like  
> exceeding maximum lock level and then coring on make test-all. Now  
> I'm not sure where the problem is. I've looked and notice this a  
> problem for several other projects as well. Probably due to the  
> poor nature of FreeBSD's pthread implementation and libthr's  
> incomplete implementation. Possible workarounds might include using  
> linux-threads. Which is what mysql recommended not too long ago.

For performance only.  Now linux-threads give worse performance than  
libthr for MySQL.

-- 
Eric Hodel - drbrain segment7.net - http://blog.segment7.net

I LIT YOUR GEM ON FIRE!


--
ML: yarv-dev-en quickml.atdot.net
Info: http://www.atdot.net/~ko1/quickml