Re: Crash in foreach_window_1 (delete-frame)

[Richard Stallman]

    *** line 6818***      if (!NILP (w->hchild))

If optimization is not confusing this somehow, then corruption of
W->hchild would not be sufficient to cause a crash there.
It would require an invalid value of W itself.

It is possible that the invalid value came out of the hchild
field in a window handled in an outer call to foreach_window_1.

Something must have been wrong with the window structure,
because a window which is an hchild should never have an hchild.
Horizontal and vertical splittings are supposed to alternate;
the only exception is that the initial window has a vertical
sibling (the minibuffer window) and may be split vertically as well.