Segfault in Emacs 23 when attempting to insert a negative integer (was: Re: infloop in skeleton-insert)

[Lawrence Mitchell]

Kenichi Handa wrote:

> In article <address@hidden>, Nick Roberts <address@hidden> writes:

>> It doesn't on Emacs 22:

>> (char-or-string-p -4)
>> t

>> and if it does on Emacs 23 then I think that must be the bug.

> Emacs 23 surely returns nil in that case.  I think the
> behaviour of Emacs 22 is a bug (or at least very confusing).

> Don't people think OBJ can be safely used as an argument of
> a function that expects a character (e.g. insert,
> char-to-string) if (char-or-string-p OBJ) is true?


This also throws up a further problem in the Emacs 23 branch:

emacs -Q 

M-: (insert -1) RET

resulting in a segmentation fault.  The problem appears to lie in
the code-path to char_string(), the code does not seem to check to see
if the "character" to be inserted is valid: specifically, negative
arguments cause the stack to overflow (I think).

char_string(c, p) calls CHAR_STRING(c, p), which, since c is negative,
calls char_string(c, p) as a fallback.

I'm not sure how to go about fixing this, the naive approach of
calling CHECK_CHARACTER at the beginning of char_string results in a
segfault when building.

[...]

Cheers,

Lawrence
-- 
Lawrence Mitchell <address@hidden>