Crash in jit-ucs-charset handling

[Aidan Kehoe]

 Ar an séiú lá de mí na Nollaig, scríobh Stephen J. Turnbull: 

 > I was running a git diff in a shell buffer.
 > 
 > Any ideas, Aidan?

On the plus side; it’s an assertion failure, so it’s an error on my part
that won’t lie undiscovered and intermittent for years. I’ve seen it on my
work Cygwin build with ISO8859-7 as the corresponding character set, so that
makes it more likely I can work out what’s causing it. 

On the minus side; I can’t reproduce it on NetBSD, even with what should be
an identical build to yours, so I can’t easily investigate further ... hmm,
actually, wait. Is it possible for Fcharset_matches_specifier_tag_set_p to
be called recursively? What’s the full C backtrace?

 > (gdb) up 4
 > #4  0x0024d520 in charset_matches_specifier_tag_set_p (charset={...}, tag_set={...}, stage=initial) at /Users/steve/Software/XEmacs/git-staging/src/specifier.c:1003
 > 1003          assert (!NILP(XVECTOR_DATA
 > (gdb) pobj charset
 > $1 = (Lisp_Charset *) 0x29c1620
 > $2 = {header = {lheader = {type = 17, mark = 0, c_readonly = 0, lisp_readonly = 0, uid = 143917}, next = 0x29c1600, uid = 143918, free = 0}, id = 207, name = {...}, doc_string = {...}, registries = {...}, short_name = {...}, long_name = {...}, reverse_direction_charset = {...}, ccl_program = {...}, to_unicode_table = 0x29c16a0, from_unicode_table = 0x2331c00, from_unicode_levels = 4, final = 48 '0', rep_bytes = 4, columns = 1, direction = 0, type = 3, dimension = 2, chars = 96, graphic = 0, encode_as_utf_8 = 1, temporary = 0}
 > (gdb) pobj $2.name
 > $3 = (Lisp_Symbol *) 0x21d02e4
 > $4 = {lheader = {type = 4, mark = 0, c_readonly = 0, lisp_readonly = 0, uid = 143900}, next = 0x37c314, name = {...}, value = {...}, function = {...}, plist = {...}}
 > Symbol name: jit-ucs-charset-0

-- 
Santa Maradona, priez pour moi!