Re: "Official" binary of Emacs-24.4 for MS-Windows

[Eli Zaretskii <eliz@gnu.org>]

> Date: Sat, 25 Oct 2014 13:08:38 +0200
> From: Dani Moncayo <dmoncayo@gmail.com>
> Cc: Emacs development discussions <emacs-devel@gnu.org>
> 
> I've tried (for the first time) the "make install-strip" target from a
> trunk build, and I've noticed something strange: the size of the
> resulting stripped "emacs.exe" is ~ 16 MB, whereas the same file for
> the 24.4 release (also stripped) is ~ 9 MB.
> 
> Such big difference is strange, isn't it?

I think it's expected.  You will see the same 7MB difference between
the unstripped src/emacs.exe in the emacs-24 branch vs the trunk.

The reason is that on the trunk we now store the memory allocated
during dumping in a static array, instead of a special section of
emacs.exe.  That static array is larger than is strictly needed for an
Emacs dumped after all Lisp files were compiled to *.elc, but we keep
the larger size because it is needed during bootstrap, when the
preloaded Lisp files are loaded in their source form.  We didn't yet
find a way to reduce the size of the array when we build a
non-bootstrap Emacs.

It would be nice to find a solution to this, but it's not a bug.

> Also, I've noticed this suspicious errors in the output of "make
> install-strip", which don't appear in a "make install" from the same
> build:

Looks like some redirection snafu: for some reason, install-info is
called with arguments taken from Make messages that announce entering
and leaving directories.  Maybe it comes from this line:

>    info_misc=`MAKELEVEL=0 make -s -C doc/misc echo-info`; \

Looks like MAKELEVEL=0 doesn't work with your Make.  Try adding the
"--no-print-directory" switch to the "make" command, and see if that
helps.

This sounds like bug #13962, but since we now require GNU Make on the
trunk, I think it's OK to use --no-print-directory instead of the
MAKELEVEL=0 trick.