From: Nathaniel Talbott <nathaniel@...>
Date: 2003-12-12T02:59:19+09:00
Subject: Re: --enable-pthread broken?

On Dec 11, 2003, at 11:49, Yukihiro Matsumoto wrote:

> In message "--enable-pthread broken?"
>     on 03/12/12, Nathaniel Talbott <nathaniel@talbott.ws> writes:
>
> |[ruby-talk: 87759] and the surrounding thread seem to indicate that
> |--enable-pthread is broken in latest CVS. Will it be fixed before 
> 1.8.1
> |final? Or is --enable-pthread just not a viable option for the
> |foreseeable future?
>
> How often do you have crash?  I had two out of many tries.

Every time I've run the test programs with --enable-pthread, one or the 
other of them has crashed (always the server except for once). Usually 
it happens within five minutes.

I don't know if this info is useful or not, but I'm running this on 
Debian stable (with some limited, unrelated upgrades to Debian 
testing).


> This might be caused by mixing longjmp and pthread, but I haven't able
> to confirm it yet.

That's my guess, too... I wish I could help track it down, but I'm 
afraid I don't know enough of the requisite black magic :-/


> Can you try the attached patch?
> At least I had no crash after this modify.

I tried it against the latest Ruby CVS, and it still crashes. It seemed 
to take (subjectively) longer to crash, though. Interestingly, 2 out of 
3 runs gave me this instead of a segfault:

   ntalbott@proxytest:~/tmp/ruby-build$ bin/ruby ~/sandbox/server.rb
   Illegal instruction
   ntalbott@proxytest:~/tmp/ruby-build$


Thanks,


Nathaniel

<:((><