|
Administrator
|
On Wednesday, August 25, 2010 07:02:16 Steve vaughan wrote:
> Sven,
>
> Hopefully you've received a pull request for the JOGL changes. I went ahead
> and introduced the focus fix that was holding up my test runs. I'm not sure
> how to set up a unit test for this particular case, since it results in a
> deadlock with the AWT EDT in native code. I'll identify the specific test
> and get back to you with the details.
I see ..
AWT EDT deadlock
==================
You cannot make an invokeLater() here since we have to guarantee
that the AWT focus request is finished _before_ we proceed with the
native focus request.
invokeLater() would allow to 'overwrite' the native focus request,
it makes the sequence non deterministik.
So I am afraid we need to get to the grounds of it .. even though I can see/assume
the possibilities of a deadlock - however, a stacktrace would be very appreciated.
Best produced with jstack, while it's hanging.
To let it hang 'a little longer', just increase the value of
com.jogamp.nativewindow.impl.RecursiveToolkitLock.timeout.
So a reproduceable test case for this issue would be great,
best if it's reproducable on GNU/Linux.
NEWT EDT lifecycle, Display, etc
=================================
A seperate test case (class file) would be appreciated.
I guess we should start a new test package with this one.
>
> BTW, I didn't need to make any SSH adjustment for git. I updated my Cygwin
> setup this evening and picked up an update to OpenSSH. Either that reduced
> the likelihood of the problem, or I just got lucky.
Good to know, thx.
>
> Let me know if you'd like some help or if you just want to discuss options.
Sure .. I allowed myself to xpost this reply to our forum as well.
Cheers, Sven
|
Locked
