AppServer crashes in SetCursor

[fishpond]

Especially when using Java applications (in x86_gcc2h), I encounter app_server crashes quite often. The backtrace always shows the crash to occur at SetCursor+0x85.

KERN: stack trace, current PC 0xfcae81 SetCursor7DesktopP12ServerCursor + 0x85: KERN: (0x727206a8) 0xfeb0f9 _DispatchMessage9ServerApplRQ28BPrivate12LinkReceiver + 0x19f1 KERN: (0x72721418) 0xff0b84 _MessageLooper9ServerApp + 0x1b0 KERN: (0x727214d8) 0xfe13c4 _message_thread13MessageLooperPv + 0x28 KERN: (0x72721508) 0x5afa9b thread_entry + 0x23

[waddlesplash]

Since this is an app_server crash, you can save a debug report by typing save-report at the prompt. It should be on your desktop after you reboot.

[fishpond]

I'll do. Now I've created the ticket, it looks like Mr. Murphy is taking care - so no more crashes so far. So we need to wait a bit more. If you want to close the ticket, feel free, I'll reopen it with a debug report attached as soon as I have it.

[fishpond]

Debug Report

[fishpond]

Finally crashed again, after quite some time - so I've now attached the debug report.

[smallstepforman]

I have a similar crash, in Desktop::SetCursor + 0x8c. Was running QtCreator at the time.

bt shows the following:

Desktop::SetCursor(ServerCursor*) +0x8c ServerApp::_DispatchMessage(int, BPrivate::LinkReceiver&) +0x6134 ServerApp::_MessageLooper() + 0x139

[korli]

which hrev?

[smallstepforman]

Reply to korli. hrev54679 (27 Oct 2020)

[waddlesplash]

I have an app_server crash here which isn't in SetCursor, just in MessageLooper, but did occur while trying to set a cursor. Unfortunately the debug report does not reveal much. This however is not a NULL dereference but rather an attempted call to a bogus pointer location.

[waddlesplash]

It's 100% reproducible for me, actually.

[waddlesplash]

I guess this is related to bitmap cursor handling which long postdates this ticket, though.

[waddlesplash]

This may be fixed after hrev57896.