Re: Mingw build failed

[Ranjit Mathew <rmathew at hotmail dot com>]

Jeff Sturm wrote:
> Any insight on the gctest failures?  It doesn't appear to me that the
> primordial thread is registered, and the obvious fix didn't work.
No. Where can one find this beast?

I'm only now getting a hang of the Win32 threads implementation
especially after seeing the SignalObjectAndWait( ) failure
on Win95 and the Object.wait( ) error reported by Erik - I found
out (from a comment in the code) that the implementation for
the condition variable support is modelled very closely on:

    http://www.cs.wustl.edu/~schmidt/win32-cv-1.html

To eliminate the reliance on the SignalObjectAndWait( ) function
which is not there on Win95 and seems to be a dummy stub on
Win98, I guess we can use the "Generation Count" solution listed
in the paper, though it seems to suffer a teeny-weeny bit
from fairness and correctness problems.

Or perhaps we can implement the "Specific Notification"
solution given in a follow-up paper to the above:

    http://www.cs.wustl.edu/~schmidt/win32-cv-2.html

This seems to be a slightly less efficient solution
than the SignalObjectAndWait( ) solution, but at least
is portable across all Win32 implementations.

What do the Magi say?

Ranjit.

--
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/