Applications stop in the middle of start process

[stippi]

I've only seen this in recent revisions (currently running hrev28334), but I cannot say when it started. Sometimes, when I launch applications, they get a Deskbar entry but open no window. This also affects very simple applications, like a preflet. Printing the stack crawl in KDL for the application thread (in my case here the only thread of the application yet) gives this:

<kernel_x86> context_switch(thread*: 0x81a13000, thread*: 0x80110bf0) + 0x003c
<kernel_x86> simple_reschedule() + 0x029f
<kernel_x86>:switch_sem_etc + 0x0319
<kernel_x86>:acquire_sem_etc + 0x0026
<kernel_x86>:_get_port_message_info_etc + 0x012e
<kernel_x86>:port_buffer_size_etc + 0x0025
<kernel_x86>:user_port_buffer_size_etc + 0x008d
<kernel_x86>:handle_syscall + 0x00af
user iframe at 0x83795fa8 (end = 0x83796000)
 eax 0xc3           ebx 0x5112fc    ecx 0x3f        edx 0xbd02f
 esi 0xffffffff     edi 0x7fffffff  ebp 0x7ffeebcc  esp 0x83795fdc
 eip 0xffff0102  eflags 0x216  user esp 0x7ffeeba0
 vector: 0x63, error code: 0x0
<commpage>:commpage_syscall + 0x0002
<libbe.so>:__c1__Q38BPrivate11BLooperList12FindPortPredRQ38BPrivate11BLooperList10LooperData (nearest) + 0x015b
[*** READ FAULT at 0xffffffff, pc: 0x800d7f83 ***]

I have no idea what recent change could have caused this. Looks like the thread is waiting for a reply, but there were no app_server or registrar changes recently, were there?

[stippi]

I can add the information that the particular application remains not "startable", even after a reboot. So possibly I have a corrupt image, but since I saw this problem across two different computers, I didn't think it would be that. Will keep you posted.

[stippi]

Re-initializing and re-populating the image did not work. jam clean and removing the generated/attributes folder and then rebuilding everything from scratch did not work.

I saw this problem with ActivityMonitor on my desktop machine and with a different application (the forthcoming TouchPad preflet) on my notebook. But now it turns out, ActivityMonitor on the notebook is also affected. In the case of ActivityMonitor, the only threads running are the application thread and the _BMediaRoster_ thread.

[anevilyak]

Someone reported seeing that on VMWare as well. I'm not seeing it on my desktop though. You're multicore right Stephan? Does disabling SMP or turning off a core in ProcessController and then trying to start the app fix it?

[stippi]

I don't remember what the fix was, but I fixed it, didn't I?