why keep generated lisp/international/uni-*.el in the repo?

why keep generated lisp/international/uni-*.el in the repo?

[Glenn Morris]

Why do we keep the generated files lisp/internation/uni-*.el, charprop.el
in the repo, rather than generating them during bootstrap?

Is it just that "they don't change very often, and it makes bootstrap a
tiny bit faster"?

On my machine, it takes me ~ 5 seconds to generate them;
compared to ~ 5 mins for a bootstrap.
It doesn't seem worth keeping them for that.

This is all we need to change to remove them from the repo:

=== modified file 'admin/unidata/Makefile.in'
*** admin/unidata/Makefile.in	2013-10-30 16:45:51 +0000
--- admin/unidata/Makefile.in	2013-11-27 08:56:18 +0000
***************
*** 64,66 ****
--- 64,70 ----
  	-rm -f ./Makefile
  
  maintainer-clean: distclean
+ 	if test -f ${DSTDIR}/charprop.el; then \
+ 	  (cd ${DSTDIR} && rm -f `sed -n 's/^;; FILE: //p' < charprop.el`); \
+ 	  rm -f ${DSTDIR}/charprop.el; \
+ 	fi

=== modified file 'src/Makefile.in'
*** src/Makefile.in	2013-11-27 06:15:06 +0000
--- src/Makefile.in	2013-11-27 08:50:26 +0000
***************
*** 417,429 ****
  $(leimdir)/leim-list.el: bootstrap-emacs$(EXEEXT)
  	cd ../leim && $(MAKE) $(MFLAGS) leim-list.el EMACS="$(bootstrap_exe)"
  
  ## The dumped Emacs is as functional and more efficient than
  ## bootstrap-emacs, so we replace the latter with the former.
  ## Strictly speaking, emacs does not depend directly on all of $lisp,
  ## since not all pieces are used on all platforms.  But DOC depends
  ## on all of $lisp, and emacs depends on DOC, so it is ok to use $lisp here.
  emacs$(EXEEXT): temacs$(EXEEXT) $(ADDSECTION) \
!                 $(etc)/DOC $(lisp) $(leimdir)/leim-list.el
  	if test "$(CANNOT_DUMP)" = "yes"; then \
  	  rm -f emacs$(EXEEXT); \
  	  ln temacs$(EXEEXT) emacs$(EXEEXT); \
--- 417,433 ----
  $(leimdir)/leim-list.el: bootstrap-emacs$(EXEEXT)
  	cd ../leim && $(MAKE) $(MFLAGS) leim-list.el EMACS="$(bootstrap_exe)"
  
+ $(lispsource)/international/charprop.el: bootstrap-emacs$(EXEEXT)
+ 	cd ../admin/unidata && $(MAKE) $(MFLAGS) all EMACS="../$(bootstrap_exe)"
+ 
  ## The dumped Emacs is as functional and more efficient than
  ## bootstrap-emacs, so we replace the latter with the former.
  ## Strictly speaking, emacs does not depend directly on all of $lisp,
  ## since not all pieces are used on all platforms.  But DOC depends
  ## on all of $lisp, and emacs depends on DOC, so it is ok to use $lisp here.
  emacs$(EXEEXT): temacs$(EXEEXT) $(ADDSECTION) \
!                 $(etc)/DOC $(lisp) $(leimdir)/leim-list.el \
!                 $(lispsource)/international/charprop.el
  	if test "$(CANNOT_DUMP)" = "yes"; then \
  	  rm -f emacs$(EXEEXT); \
  	  ln temacs$(EXEEXT) emacs$(EXEEXT); \

Re: why keep generated lisp/international/uni-*.el in the repo?

[Andy Moreton]

On Wed 27 Nov 2013, Glenn Morris wrote:

> Why do we keep the generated files lisp/internation/uni-*.el, charprop.el
> in the repo, rather than generating them during bootstrap?
>
> Is it just that "they don't change very often, and it makes bootstrap a
> tiny bit faster"?
>
> On my machine, it takes me ~ 5 seconds to generate them;
> compared to ~ 5 mins for a bootstrap.
> It doesn't seem worth keeping them for that.

If the files only chnage when a new version of Unicode appears, then
keeping things as they are seems far more sensible. Why waste power and
the time of users if this stuff almost never changes ?

On older hardware a boostrap can take considerably longer, so I don't
think this change would be useful.

    AndyM

Re: why keep generated lisp/international/uni-*.el in the repo?

[Glenn Morris]

Actually, I'll just do what we do with ja-dic.el:
Remove from repo, but not delete even in a bootstrap- or maintainer-clean.
Then the generation becomes a one time cost, and everybody wins.

Re: why keep generated lisp/international/uni-*.el in the repo?

[Stefan Monnier]

> Why do we keep the generated files lisp/internation/uni-*.el, charprop.el
> in the repo, rather than generating them during bootstrap?

Good question.

> Is it just that "they don't change very often, and it makes bootstrap a
> tiny bit faster"?

Maybe it's only historical (the scripts to build them were run in an
ad-hoc manner).  Or maybe they rely on tools which may not always be
available?

> On my machine, it takes me ~ 5 seconds to generate them;
> compared to ~ 5 mins for a bootstrap.
> It doesn't seem worth keeping them for that.

As a user of slow machines where bootstrap is rather in the half-hour
area, I think I wouldn't care about an extra 30 seconds, indeed.


        Stefan