two darwin issues

Wed Feb 25 04:58:00 GMT 2004

Hi Geoff,
Geoff Strom wrote:
> I've built recent cvs versions of 3.4 on both Jaguar and Panther and can happily report that both JNI and threads seem to work fine.
>> However I am seeing two specific failures that I'm not able to adequately debug.
>> The first is running SWT programs. I can compile a SWT based program fine. When running it certainly looks like it is linking to the .dylib correctly, but it always dies after some delay with a "Bus Error" (condensed crash log #1 below).
>> The second issue seems like a regression. I can build and run Mozilla's Rhino interpreter and it works great for pure javascript. Any time I attempt to use LiveConnect and load Java objects, it dies immediately (condensed crash log #2 below). I swear this worked for me before when I was working with GCC 3.3.1 (FSF).
>> Could either of these be affected by gc? Should I try updating to 6.3alpha4?
>> How can I help find the real issues here?

Just a fyi, darwin is missing an important part to handle NPE correctly. 
I sent a preliminary patch around the 30.1 to gcc-patches, it 
implements the MD_FALLBACK_FRAMESTATE_FOR macro which is neccessary to 
unwind in case of a NPE. (Null Pointer Exception).
Also a patch to libjava signal handling is necessary.
And second, between 3.3.x and 3.4.x fsf gcc on darwin got the dylib 
implementation. So it's no longer a static library .....
These things maybe related to your failures you see.
We hope to get these things sorted out soon.
I was able to run some awt gtk stuff, yes some windows without swing.
Andreas