bug#4789: leim-list.el needs to be included in the dumped images

bug#4789: leim-list.el needs to be included in the dumped images

[Dan Nicolaescu]

leim-list.el is loaded at run time even in batch mode, it's a waste to
run that code all the time, it should be included in the dumped image.

Also see:
http://permalink.gmane.org/gmane.emacs.devel/116284

To not waste space, register-input-method needs to be changed to
purecopy all it's string arguments.

bug#4881: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Dan Nicolaescu]

emacs/leim/leim-list.el is big and it gets loaded every time emacs is
started, even in batch mode.  It would be more efficient to put it in
the dumped image, it will be stored in read-only memory that way.

For that emacs/leim/leim-list.el needs to be created before dumping, but
the emacs/leim directory is built with "emacs", it should be built with
"bootstrap-emacs".

See also: http://permalink.gmane.org/gmane.emacs.devel/116284

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Lars Magne Ingebrigtsen]

Dan Nicolaescu <dann@ics.uci.edu> writes:

> emacs/leim/leim-list.el is big and it gets loaded every time emacs is
> started, even in batch mode.  It would be more efficient to put it in
> the dumped image, it will be stored in read-only memory that way.

If I do

strace emacs -Q 2>&1 | grep leim

I get nothing.  It doesn't seem to load that file in Emacs 24, so I'm
closing this bug.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Glenn Morris]

Lars Magne Ingebrigtsen wrote:

> strace emacs -Q 2>&1 | grep leim
>
> I get nothing.

I get loads of matches, and I don't think anything has changed in this
area, so I'll reopen this.

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Lars Magne Ingebrigtsen]

Glenn Morris <rgm@gnu.org> writes:

> Lars Magne Ingebrigtsen wrote:
>
>> strace emacs -Q 2>&1 | grep leim
>>
>> I get nothing.
>
> I get loads of matches, and I don't think anything has changed in this
> area, so I'll reopen this.

Do you know why your Emacs loads leim?  Is there a build option or
something?

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Glenn Morris]

Lars Magne Ingebrigtsen wrote:

> Do you know why your Emacs loads leim?  Is there a build option or
> something?

Nope. Do you know why yours doesn't? :)
It's part of normal-top-level in startup.el.
(It's not the kind of thing that should be changed for Emacs 24.1 at
this stage.)

http://lists.gnu.org/archive/html/emacs-devel/2009-10/msg00502.html

points out that there still has to be _some_ search at startup, for
people with their own input methods, and suggests just not searching
certain "obviously wrong" load-path dirs (lisp/progmodes, etc). That
does not preclude dumping the standard leim/leim-list.el though.

Oh, I thought this seemed familiar:

http://debbugs.gnu.org/cgi/bugreport.cgi?bug=910

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Lars Magne Ingebrigtsen]

Glenn Morris <rgm@gnu.org> writes:

> Nope. Do you know why yours doesn't? :)

I don't even know what leim is.  :-)

> It's part of normal-top-level in startup.el.
> (It's not the kind of thing that should be changed for Emacs 24.1 at
> this stage.)
>
> http://lists.gnu.org/archive/html/emacs-devel/2009-10/msg00502.html
>
> points out that there still has to be _some_ search at startup, for
> people with their own input methods, and suggests just not searching
> certain "obviously wrong" load-path dirs (lisp/progmodes, etc). That
> does not preclude dumping the standard leim/leim-list.el though.

Right.  Was there any consensus that dumping with leim-list would be a
good idea?  I read the thread, and I didn't really see any conclusions.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Andreas Schwab]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> I don't even know what leim is.  :-)

leim/README

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Glenn Morris]

Lars Magne Ingebrigtsen wrote:

> Right.  Was there any consensus that dumping with leim-list would be a
> good idea?  I read the thread, and I didn't really see any conclusions.

Don't know. FWIW, I think leim/leim-list should be dumped (and therefore
would need to be built with bootstrap-emacs), and normal-top-level
should ignore those elements of load-path that are part of Emacs and so
won't contain a leim-list. I doubt it will have much of a measurable
effect though, so it's not a high priority to change it.

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Dan Nicolaescu]

Glenn Morris <rgm@gnu.org> writes:

> Lars Magne Ingebrigtsen wrote:
>
>> Right.  Was there any consensus that dumping with leim-list would be a
>> good idea?  I read the thread, and I didn't really see any conclusions.
>
> Don't know. FWIW, I think leim/leim-list should be dumped (and therefore
> would need to be built with bootstrap-emacs), and normal-top-level

Or just check in the generated leim-list.el, it almost never changes.
We do something similar with ldefs-boot.el... 

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Glenn Morris]

Glenn Morris wrote:

> FWIW, I think leim/leim-list should be dumped (and therefore would
> need to be built with bootstrap-emacs), and normal-top-level should
> ignore those elements of load-path that are part of Emacs and so won't
> contain a leim-list.

The normal-top-level part is now implemented in the trunk.

On GNU/Linux, leim-list.el is now generated with bootstrap-emacs
(see 2012-04-09 change to src/Makefile.in).

If Someone (TM) will kindly implement the same thing for MS Windows, we
will be able to load leim-list.el in loadup.el and close this
(triplicate) report.

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Date: Mon, 09 Apr 2012 03:20:10 -0400
> 
> On GNU/Linux, leim-list.el is now generated with bootstrap-emacs
> (see 2012-04-09 change to src/Makefile.in).

Is there any special reason to do that with bootstrap-emacs and not
with a "normal" emacs?  I'm asking because there's no bootstrap-emacs
on Windows.

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Glenn Morris]

Eli Zaretskii wrote:

> Is there any special reason to do that with bootstrap-emacs and not
> with a "normal" emacs? 

So that leim-list can be included in the dumped emacs executable.

> I'm asking because there's no bootstrap-emacs on Windows.

Oh. Then I guess I mean whatever executable is used to compile the lisp
files from loadup.el on MS Windows.

Basically, I want to load leim-list.el in loadup.el, so it needs to be
generated before loadup.el is read. Similar to loaddefs.el and
ldefs-boot.el, except with no leim-boot.el being needed (I guess; I see
I haven't thought this bit through!).

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Cc: 4789@debbugs.gnu.org
> Date: Mon, 09 Apr 2012 03:59:42 -0400
> 
> Eli Zaretskii wrote:
> 
> > Is there any special reason to do that with bootstrap-emacs and not
> > with a "normal" emacs? 
> 
> So that leim-list can be included in the dumped emacs executable.

But you need to do this again whenever Emacs is rebuilt, not just at
bootstrap time.  Right?  Are you relying on the fact that on Posix
hosts bootstrap-emacs and emacs are (currently) the same executable,
and that bootstrap-emacs is generated even when no bootstrap is
necessary, like when building a released tarball?

> > I'm asking because there's no bootstrap-emacs on Windows.
> 
> Oh. Then I guess I mean whatever executable is used to compile the lisp
> files from loadup.el on MS Windows.

So you just need _any_ emacs executable capable of code that generates
leim-list.el, is that right?

> Basically, I want to load leim-list.el in loadup.el, so it needs to be
> generated before loadup.el is read. Similar to loaddefs.el and
> ldefs-boot.el, except with no leim-boot.el being needed (I guess; I see
> I haven't thought this bit through!).

Well, the actual intent of my question was to see if we need to
discuss the design of this some more ;-)

Thanks.

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Glenn Morris]

Eli Zaretskii wrote:

> But you need to do this again whenever Emacs is rebuilt, not just at
> bootstrap time.  Right?  Are you relying on the fact that on Posix
> hosts bootstrap-emacs and emacs are (currently) the same executable,
> and that bootstrap-emacs is generated even when no bootstrap is
> necessary, like when building a released tarball?

Indeed I need to think about this more... My initial thought is that it
should be treated like loaddefs.el (and finder-inf.el, cus-load.el,
etc). All these are pre-built in the release tarfile, so maybe
leim-list.el should be too. Now that I think about it, it's suprising
that it isn't already.

> So you just need _any_ emacs executable capable of code that generates
> leim-list.el, is that right?

Right. But when bootstrapping on MS Windows, there must still be two
versions of the emacs binary, right? One that is called early on and
loads uncompiled Lisp files (whether or not it is called
"bootstrap-emacs"), and is used to generate the compiled Lisp files that
are loaded by the final, "proper" Emacs? If so, I mean the first one.

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Cc: 4789@debbugs.gnu.org
> Date: Mon, 09 Apr 2012 12:34:10 -0400
> 
> Eli Zaretskii wrote:
> 
> > But you need to do this again whenever Emacs is rebuilt, not just at
> > bootstrap time.  Right?  Are you relying on the fact that on Posix
> > hosts bootstrap-emacs and emacs are (currently) the same executable,
> > and that bootstrap-emacs is generated even when no bootstrap is
> > necessary, like when building a released tarball?
> 
> Indeed I need to think about this more... My initial thought is that it
> should be treated like loaddefs.el (and finder-inf.el, cus-load.el,
> etc). All these are pre-built in the release tarfile, so maybe
> leim-list.el should be too. Now that I think about it, it's suprising
> that it isn't already.

Well, that goes hand in hand with the fact that the *.el files under
leim/ are not byte-compiled in the tarball.  Basically, a remnant from
ye-old days when Leim was a separate package.  I think we should
compile all the *.el files under leim/ as well, but Stefan once told
me something that I interpreted to mean he would rather go in the
opposite direction -- let the build process compile all the *.el files
under lisp/ and ship them uncompiled in the tarball.  Maybe I
misunderstood.

> > So you just need _any_ emacs executable capable of code that generates
> > leim-list.el, is that right?
> 
> Right. But when bootstrapping on MS Windows, there must still be two
> versions of the emacs binary, right? One that is called early on and
> loads uncompiled Lisp files (whether or not it is called
> "bootstrap-emacs"), and is used to generate the compiled Lisp files that
> are loaded by the final, "proper" Emacs?

Yes.

> If so, I mean the first one.

Yes, that part is clear.

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Glenn Morris]

Eli Zaretskii wrote:

> Well, that goes hand in hand with the fact that the *.el files under
> leim/ are not byte-compiled in the tarball.  Basically, a remnant from
> ye-old days when Leim was a separate package.  I think we should
> compile all the *.el files under leim/ as well,

Me too. I'll look into this.

> but Stefan once told me something that I interpreted to mean he would
> rather go in the opposite direction -- let the build process compile
> all the *.el files under lisp/ and ship them uncompiled in the
> tarball. Maybe I misunderstood.

I think you are referring to http://debbugs.gnu.org/2151 .

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Cc: 4789@debbugs.gnu.org
> Date: Mon, 09 Apr 2012 13:30:45 -0400
> 
> > but Stefan once told me something that I interpreted to mean he would
> > rather go in the opposite direction -- let the build process compile
> > all the *.el files under lisp/ and ship them uncompiled in the
> > tarball. Maybe I misunderstood.
> 
> I think you are referring to http://debbugs.gnu.org/2151 .

Yes.  More specifically, this part of
http://debbugs.gnu.org/cgi/bugreport.cgi?bug=2151#35:

  OTOH, I feel like it's fundamentally wrong to provide pre-built
  files in our tarball.

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Stefan Monnier]

> compile all the *.el files under leim/ as well, but Stefan once told
> me something that I interpreted to mean he would rather go in the
> opposite direction -- let the build process compile all the *.el files
> under lisp/ and ship them uncompiled in the tarball.
> Maybe I misunderstood.

What I said/meant was that we should reduce the difference between
"build from tarball" and "build from Bzr checkout".  So, yes, I'd rather
we don't distribute the .elc files in the tarball.  But it's a low
priority goal.


        Stefan

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: Glenn Morris <rgm@gnu.org>,  4789@debbugs.gnu.org
> Date: Mon, 09 Apr 2012 21:52:59 -0400
> 
> > compile all the *.el files under leim/ as well, but Stefan once told
> > me something that I interpreted to mean he would rather go in the
> > opposite direction -- let the build process compile all the *.el files
> > under lisp/ and ship them uncompiled in the tarball.
> > Maybe I misunderstood.
> 
> What I said/meant was that we should reduce the difference between
> "build from tarball" and "build from Bzr checkout".  So, yes, I'd rather
> we don't distribute the .elc files in the tarball.  But it's a low
> priority goal.

Does it mean you will be against distributing Leim packages in
byte-compiled form?

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Stefan Monnier]

>> > compile all the *.el files under leim/ as well, but Stefan once told
>> > me something that I interpreted to mean he would rather go in the
>> > opposite direction -- let the build process compile all the *.el files
>> > under lisp/ and ship them uncompiled in the tarball.
>> > Maybe I misunderstood.
>> What I said/meant was that we should reduce the difference between
>> "build from tarball" and "build from Bzr checkout".  So, yes, I'd rather
>> we don't distribute the .elc files in the tarball.  But it's a low
>> priority goal.
> Does it mean you will be against distributing Leim packages in
> byte-compiled form?

I'm OK with it if it simplifies the code (by re-using the code used for
the `lisp' directory), but otherwise I'm generally not in favor, indeed.


        Stefan

bug#4789: use bootstrap-emacs to build emacs/leim directory, load emacs/leim/leim-list.el to loadup.el

[Glenn Morris]

Version: 24.4

Done.