bug#21688: [Emacs-diffs] emacs-25 8ff888a: Improve fix for regex reentrancy abort

bug#21688: [Emacs-diffs] emacs-25 8ff888a: Improve fix for regex reentrancy abort

[Paul Eggert]

John Wiegley wrote:
>>>>>> Ken Brown<kbrown@cornell.edu>  writes:
>>> >>branch: emacs-25
>>> >>commit 8ff888a07d0500f1274723086469191aade0a33e
>> >The message-mode-propertize test fails after this commit.
> Let's make sure all tests are run prior to any commits on the emacs-25 branch.

I briefly looked into getting that test working, but couldn't figure it out 
quickly, so I have reverted the fix for the regex reentrancy abort, in emacs-25 
commit 8e84823f3f06714d0668060d62456cf7e3100e6c.  This means Bug#21688 is no 
longer fixed, i.e., Emacs can dump core when aligning Perl code.  I'll reopen 
Bug#21688.

Perhaps Stefan or someone else will be able to knock a few cycles free to fix 
this. I'm afraid I've been spending most of my time elsewhere in Emacs recently, 
what with the new module code and all.

bug#21688: [Emacs-diffs] emacs-25 8ff888a: Improve fix for regex reentrancy abort

[Stefan Monnier]

> Perhaps Stefan or someone else will be able to knock a few cycles free to
> fix this. I'm afraid I've been spending most of my time elsewhere in Emacs
> recently, what with the new module code and all.

My crystal ball tells me that maybe it's just the following hunk which
is to blame:

   diff --git a/src/syntax.c b/src/syntax.c
   index 5b0ec6d..f939a76 100644
   --- a/src/syntax.c
   +++ b/src/syntax.c
   @@ -514,7 +514,6 @@ update_syntax_table_forward (ptrdiff_t charpos, bool init,
        {
          eassert (NILP (object));
          eassert (charpos >= gl_state.e_property);
   -      parse_sexp_propertize (charpos);
        }
      else
        {

I think this hunk is just a mistake that was introduced during the
"back&forth" that took place when you added and then removed the "bool
propertize" argument.


        Stefan