[regression] thread behavior: interruption of load and render

[Bill Haneman <Bi...@ireland.sun.com>]

Hi:

Sorry to break this news, but there appears to be some
significant regression regarding interruption of document
loading and rendering.

In the past, when "File->Open File" was used while a document
load or render was already in progress, the resulting
load of a new document would cause the background rendering
or loading to be interrupted.  This was not only an efficiency
improvement (since the "undesired" action would be aborted)
but also prevented the onscreen display from getting out-of-sync
with the document loading part of the GUI.  Both problems now
appear (out-of-sync display, and document load/render that
continues despite being pre-empted).

I suspect that this regression crept in during some of
the recent error handling changes, but I may be wrong.
At any rate the 1.0beta release did not have this problem.

It's alse possible that this problem is not present in the
most current CVS, I have not done an update for a few
days.  My version is from Jan. 5 I believe.

I don't have time to chase this now, but if nobody has
found the problem by the end of the week I will go hunting.
If you have questions about the error handling/threading
in JSVGCanvas, please ask me before making changes!  Perhaps
I can explain why certain design decisions were made, at any
rate I'd like to discuss changes first.

A less important point: the loading process now
throws java.lang.NoClassDefFoundError, where before it
threw java.lang.ClassNotFoundException.  The latter error
was/is caught by the JSVGCanvas (it is generated when a
loading/rendering thread is interrupted, under certain
circumstances, since these processes are not smart
regarding interruption at the present time and do not
detect that they have been interrupted).  The former
behavior is a workaround which allows us to interrupt such
threads without having to "instrument" the working methods
for interrupt-checking.  There are other errors which now
get caught, sometimes, and presented to the user when a
document load/render is interrupted, when they should not
be.

-Bill

------
Bill Haneman x19279
Ireland Desktop Engineering
Sun Microsystems Ireland Ltd.

Re: [regression] thread behavior: interruption of load and render

[Thierry Kormann <Th...@sophia.inria.fr>]

> Sorry to break this news, but there appears to be some
> significant regression regarding interruption of document
> loading and rendering.

After several tries, I was not able to see your problem :(
Do you still have the same problem with the current code base?
If yes, could you email me a way to reproduce the error?

Regards,
Thierry.

-- 
Thierry Kormann
email: Thierry.Kormann@sophia.inria.fr  http://www.inria.fr/koala/tkormann/
Koala/Dyade/Bull @ INRIA - Sophia Antipolis