Re: pcp updates: more multithreaded fixes and then some

[fche@xxxxxxxxxx (Frank Ch. Eigler)]

kenj wrote:

> [...]
>> 449 - output mismatch (see 449.out.bad)
> I'm also seeing this across a handful of QA machines.
> [...]

This new commit on fche/multithread returns to the pmns.c locking
structure of a week or two ago, but using a private recursive lock
instead of libpcp.  This appears to make qa/449 reliable here.
I'd appreciate a look-see.

commit 516f3c136517edd5dc06f3738010c0a80ad609ed
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Wed May 11 12:28:24 2016 -0400

    libpcp multithreading: pmns.c switch to recursive lock
    
    Reverting and reworking the previous pmns.c lock-unnesting patch by
    using a recursive pmns_lock.  This is necessary because the internal
    dynamic callgraph within pmns.c is too DAGgy to put a normal mutex
    just at the top levels entry points, and the static state being
    protected is too diffuse just to put a normal mutex at the bottom.
    All this appears to make qa/449 happy now.


- FChE