Re: emacs doesn't compile (again)

[Roland McGrath]

> emacs doesn't compile.  Yippie.

Emacs internals have changed a lot since last time I hacked on them,
so I have less specific help here than I might have.

> The funny thing is that I can reproduce this with the normal libs
> and the libc debug libs, and in the debugger with the normal libs.
> But I can't reproduce it in the debugger with the debug libs.

Muwahahaha.  It's all part of the world-wide conspiracy to drive you insane.

There should be no difference observable at runtime between the two sets of
libs.  Both sets of libs should be identical after going through
--strip-debug.  If not, the libc build is questionable.  See how they
differ under objdump -hrd.  If it's a truly obscure situation, it could be
that gdb doing a lot of disk/memory to load up symbols perturbs the world
so as to make it not happen.  We need to know more about the fault you are
getting.

> The debug session with non-symboled libs is below.  It doesn't seem
> to make much sense, the backtrace looks a bit bogus (I don't see how
> it comes from frame #1 to #0).

Whenever you have a fault, always show "info regs" and "disas $pc".

The backtrace can indeed be bogus if the stack was in a funny state at the
crash.  You might try using -lmcheck (assuming that works with emacs on linux)
just to make sure it's not a simple memory clobberation bug in emacs that
only happens to show up under the hurd.