On Mon, 1 Jun 2009, Peter W. Draper wrote:
> Hi Mark,
>
> I've just committed a batch of changes to JNIAST to support windows, which
> after a lot of pain seems to working.
>
> The main points are:
>
> 1 - I've removed all use of env_key. There is a JNI method that gets
> you the same functionality without worrying about synchronisation
> so I've changed to that.
>
> 2 - I've added a new GrfLock synchronisation object. Does the same
> as AstLock, but for all the GRF calls that are outside the
> AstLock managed region.
>
> 3 - There's a new HAVE_PTHREAD macro that controls the inclusion of
> pthreads.h. If that's missing my "bd" implementation is used
> instead (note that this switch was less trivial than it initial
> seemed, in fact you were using pthreads even when AST wasn't,
> hence the other changes to improve all the synchronisation
> which was always being handled by pthreads).
>
> Have a look, if you want, and see how it all looks to you.
I've just got round to taking a quick look - haven't been through
it with a fine toothcomb, but seems plausible enough and I'm sure you
know what you're doing.
--
Mark Taylor Astronomical Programmer Physics, Bristol University, UK
[log in to unmask] +44-117-928-8776 http://www.star.bris.ac.uk/~mbt/
|