[Bug: 21.5-b27] font-lock-fontify-* / infinite loop

[Stephen J. Turnbull]

Wulf C. Krueger writes:

 > Ok, sorry, I must admit that I'm not really fluent in Lisp.

No need to apologize!  You grepped in the right place as Aidan points
out; and I should have taken the hint to look at `lambda' myself.

 >  SJT> You don't have to; it's a Gnus "feature" (Microsoft
 >  SJT> Windows-style), and you have to disable it (at least, that used to
 >  SJT> be the case).
 > 
 > Ah, ok, thanks. I've disabled it as you kindly described in
 > <877ivtfuw2.fsf at uwakimon.sk.tsukuba.ac.jp> but, alas, it didn't help.

Right; if Aidan is correct about where that regexp is coming from,
that particular instance isn't protected by the control variable.

 > Still trying. What I don't understand is why Adrian can't reproduce my

That's "Aidan", Adrian is a different XEmacs developer.  I only make a
point of it because *Adrian* is also very good about trying to
reproduce bugs when he's active, and uses Gnus, so *I* get confused. :-)

 > problem. Apart from some minimal additional elisps I don't use anything
 > special and the error occurs even without any of my own stuff.

There are a number of variables here.  64 bit v. 32 bit processor,
21.4 v. 21.5, compiler versions and optimizations, etc.  Aidan
probably has a 64 bit processor and is testing with the same version
as you are, but I don't recall for sure.  And if there's a compiler
bug or an inappropriate optimization, anything could happen.

I'll have to take a closer look at that backtrace you posted because
it has a regexp in it.  It doesn't look like a problematic regexp (ie,
one that matches the null string so a calling function might loop in
place), but it does contain a shy group, which we've had problems with
in the past.  I'm pretty sure that the 21.4 and 21.5 regexp engines
have diverged quite a bit, maybe the relevant fix didn't get back
ported to 21.4 or it was done incorrectly.