bug#47244: 28.0.50; SIGSEGV in long-runnning Emacs

[martin rudalics]

> Before that, I include the backtrace with your latest patch.  And here's
> the interesting bit:
>
> (gdb) p Vwindow_list
> $2 = XIL(0x55555863cae3)
[...]
> (gdb) p Vwindow_list_2
> $7 = XIL(0)

The last one should be the clue, indeed.  If it were the expected six
windows list we couldn't say much, but nil means it got reset somewhere
and never resurrected by window_list.

> Right before this I hit a breakpoint that that I had set up that, once
> again, implied that in this call of window_list() happened subsequent to
> a call to window_list() that didn't complete, somehow.
>
> Before running with block_input(), unblock_input(), I'm going to modify
> the sources to set a physical variable to one on entrance to the if
> block and set it back to zero at the exit.  At least then I can be sure
> that an unintended termination of this function is really happening and
> that it is not a debugger artifact.

Good.  Otherwise we're back at zero.

martin