On Dec 11, 2003, at 20:48, Yukihiro Matsumoto wrote:

> Nathaniel, let me ask you few more questions before start fighting.
>
>  * did --enable-pthread help your programs from blocking?

I can't say. --enable-pthread caused the segfault, so I don't think the 
app could get to the point where it would block. I have not been able 
to test my larger program with Nobu's patch and --enable-pthread yet. 
But since killer.rb has the blocking problem with the patch and 
--enable-pthread, I don't really want to test my larger program under 
load again until the problem in killer.rb is resolved.


>  * if I understand you correctly, after applying Nobu's patch, simple
>    killer.rb in [ruby-core:1938] start blocking, right?

I think killer.rb would have blocked eventually anyhow, but never got 
that far before Nobu's patch because of the segfault. So I wouldn't 
think Nobu's patch has anything to do with the blocking; rather it just 
allowed the interpreter to run long enough to hit the block condition.


>  * thus, killer.rb does blocks whether --enable-pthread is applied or
>    not, correct?

I have only had killer.rb block so far with --enable-pthread... I am 
testing it now to see if I can get it to block without 
--enable-pthread. However I have had my main app block without 
--enable-pthread. To summarize:

            |                    enable-pthread                |
            | with (unpatched) | with (patched) | without      |
---------------------------------------------------------------
  Main app  | segfault         | untested       | blocked      |
---------------------------------------------------------------
killer.rb  | segfault         | blocked        | inconclusive |

I hope this helps; please let me know if there's any more information 
you need. I'll let you know if I manage to get killer.rb to block 
without --enable-pthread.

Thanks,


Nathaniel

<:((><