Re: allocate_string_data memory corruption

["Richard M. Stallman" <rms@gnu.org>]

    > I think instead they should be disallowed in signal handlers.
    > And AFAIK they are disallowed.  But it's quite possible that some signal
    > handler does it, even though it's disallowed.

    Why, is there any disadvantage to using BLOCK_INPUT in allocate_string
    and Fcons?  It seems like a more robust solution than checking
    individual signal handlers to make sure they behave properly.

Either approach can work.  If we've been using the solution of not
allocating Lisp objects in the signal handlers, let's stick with that
for the time being.

In any case please do try the debugging statement that Stefan
suggested, so we can see if this is indeed the cause of the problem.
And it would be good to install that debugging statement permanently,
for as long as we still use this design.