bug#8623: 23.3.50; (woman "git-remote") crashes

[Paul Eggert <eggert@cs.ucla.edu>]

On 05/06/11 03:15, Eli Zaretskii wrote:
> Paul, could you please take a look at this?  Could this be due to the
> same problems with GCC 4.6 optimizations you recently fixed on the
> trunk?  If so, we may need similar changes on the release branch.

Yeowch.  I don't have easy access to a Debian unstable host,
and even if I did, I don't relish the job of looking into the
details as to why it crashed.

That aliasing patch was motivated by two things:
(1) I had a smoking gun in which GCC 4.6.0 clearly generated code
that would crash Emacs, when Emacs was compiled in debugging mode.
(2) I had the strong suspicion that other aliasing problems
were elsewhere in Emacs, even when not debugging, and that it
would be very difficult to track them down.

I'd guess that we're seeing (2) here, but *proving* that could
be a lot of work.

How about if we do this instead: (A) merge the aliasing patch into
Emacs 23, (B) see if this works for Sven; (C) if it does,
commit it into the 23 branch.  That should be less work than
proving that this is an instance of (2).