Re: SIGABRT while stress-testing with gnuclient/gnudoit

[Richard Stallman]

What I can deduce from the info in the bug report is that the frame
was dead before y-or-n-p was called.  I think the next step is to
look at what primitives were called before that.  Which was the
last one to examine SELECTED_FRAME?  At that time, the selected
frame must have been still live.

If you can get some suspicions about which primitive might have caused
the selected frame to be dead, you could put a call to SELECTED_FRAME
at entry and another just before it returns.  Then if that primitive
is responsible you will get clear proof.