Re: Emacs pretest 28.0.90 is out

[Eli Zaretskii]

> From: Po Lu <luangruo@yahoo.com>
> Cc: emacs-devel@gnu.org
> Date: Sat, 11 Dec 2021 09:06:26 +0800
> 
> >> +ifneq ($(MSDOS_OBJ),)
> >> +temacs$(EXEEXT): $(LIBXMENU) $(ALLOBJS) $(LIBEGNU_ARCHIVE) $(EMACSRES) \
> >> +  $(charsets) $(charscript) ${emoji-zwj} $(MAKE_PDUMPER_FINGERPRINT)
> >> +  $(AM_V_CCLD)$(CC) -o $@ \
> >> +    $(ALL_CFLAGS) $(TEMACS_LDFLAGS) $(LDFLAGS) \
> >> +    $(ALLOBJS) $(LIBEGNU_ARCHIVE) $(W32_RES_LINK) $(LIBES)
> >> +  $(MKDIR_P) $(etc)
> >> +else
> >>  ## We have to create $(etc) here because init_cmdargs tests its
> >>  ## existence when setting Vinstallation_directory (FIXME?).
> >>  ## This goes on to affect various things, and the emacs binary fails
> 
> > Couldn't you use the original recipe by a trivial editing (it seems
> > all you need is to replace $@.tmp with $@?), instead of adding an
> > ifdef?  Come to think of that, why not use the original recipe without
> > any changes, as it moves $@.tmp to $@ after the link succeeds?
> 
> The linker apparently doesn't output a "stub" if the filename it's
> assembling to doesn't end with ".exe", resulting in an invalid binary.

That can be remedied by an explicit "stubify foo" command.  I believe
there are examples of that already in src/Makefile.  Or you could call
the temporary file $@-tmp instead of $@.tmp, so that it has no
extension.

And in any case, even if you want not to go through a temporary file,
I'd prefer to edit the original recipe in sed1v2.inp instead of adding
a DOS-specific recipe, as the former is safer for changes on the
release branch.