Comment 4 for bug 974017

I think there are two bugs:

* The X server has to maintain the state of all logically active touches. By default, it allocates X (depending on device properties) number of touch records. When this number of touches is exhausted, say by putting lots of touches on the device, more touch records are allocated. If you touch fast enough, the server will fail to allocate fast enough to handle new touches, which come in during SIGIO context. The server then attempts to print an error message, but in a signal unsafe way, and X crashes under valgrind 100% of the time. I have not seen it crash in real life, but this is still bad and hampers investigation of the issue

* I can confirm that a different stack trace occurs when X crashes in real life, so this is likely a second issue.