Team resources might not be freed on exit (when it was killed?)

[diver]

app_server doesn't close app windows if app was killed. Steps to reproduce: Start DiskProbe and close it. Start FileTypes and set Generic file to open with DiskProbe. No go to /boot/beos/etc/Keymap, select all and press enter. You will see that diskprobe will open every Keymap (sometimes with messages " could not resolve link" or smth like that. Now try to kill DiskProbe with Vulcan death grip from Deskbar. DiskProbe will be killed, but all opened windows will be still there. Tested with rev17033 under vmware.

[axeld]

1) I don't get any error messages from DiskProbe 2) I can't reproduce the bug - all windows are closed just fine over here. (tested on real hardware)

[diver]

I will retest it right now on real hardware with latest revision. Stay tuned :)

[diver]

(In reply to comment #1)

1) I don't get any error messages from DiskProbe

I still can reproduce this one. Error message is coming from Tracker, not DiskProbe

2) I can't reproduce the bug - all windows are closed just fine over here. (tested on real hardware)

I can't either, maybe already fixed?

Should we close this bug or left for until #1 will be resolved?

[axeld]

I actually don't remember having fixed this, but I know for sure that I saw it some time ago. So even if this is not reproducible anymore the way you found, I think we should keep this one open for now.

[darkwyrm]

I've got a reproducible way for you: kill any app via the Deskbar (Ctrl+Alt+Win+Shift and click on a team). The app window goes nowhere - all that can be done is to move the zombie off to the side of the window.

[diver]

Now when i try to go to /boot/beos/etc/Keymap, select all and press enter Tracker crashes, back trace will follow, tested under vmware with r17519.

[diver]

Also while this debug window is opened a can't launch any app from Deskbar...

[diver]

back trace

[axeld]

This seems to be a kernel bug: for some reason, the team's resources aren't deleted on exit.

[axeld]

Fixed in r17652. Threads could get rescheduled too early on exit, and thus couldn't return their death stacks anymore (so their team resources never got freed, too).

[diver]

I'm sorry, but i have to reopen this bug. I just tried to do what i described in my first message and app_server crashed. Tested with r17671 under vmware. back trace will follow.

[diver]

back trace

[axeld]

Whatever this bug is, it has nothing to do with the one before. Please open a new bug for this.

[diver]

Crash mentioned in comment #12, explained in #653 should be fixed with bug 686. Tracker crash from comment #7 is now sepatated to #653.