Re: emacs doesn't compile (again)

[Roland McGrath]

> Here is a new session.  It turns out that real_morecore is 100 bytes off
> from __morecore in the dumped executable, 

This is the problem.  It ought to be resolving to a PLT address in emacs's
own .plt section, so the address is constant.  It seems to be set to an
address within libc itself.  Note that gdb can be a bit confusing when you
are trying to figure out what happened with this, because the .plt symbol
in emacs and the .text symbol in libc have the same name and I can never
figure out which thing gdb is talking about.  But you can tell from the
addresses, and you can find emacs's .plt symbol with objdump --dynamic-syms.

> Actually, a call to r_alloc_reinit in emacs.c (main) was removed in 1999 by
> RMS, and I can not find any place where it is actually called.
> So it seems that this might be the problem.  I will try to insert a call to
> r_alloc_reinit in main() and see if this leads to something funny (like a
> version of emacs which actually works).

This will probably just work around it long enough to have some other weird
problem because of the bad relocations.