parallel bootstrap failure; _prompt_? during bootstrap build

parallel bootstrap failure; _prompt_? during bootstrap build

[Jim Meyering]

I've just noticed a bootstrap failure.
I was building with make -j5:

Here's the tail of the build output.
Patch below.

        EMACSLOADPATH=/e/emacs/lisp ../src/bootstrap-emacs -batch --no-site-file --multibyte -l autoload --eval '(setq generated-autoload-file "/e/emacs/lisp/loaddefs.el")' -f batch-update-autoloads $wins
Directories: /e/emacs/lisp/. /e/emacs/lisp/./emulation /e/emacs/lisp/./play /e/emacs/lisp/./mail /e/emacs/lisp/./url /e/emacs/lisp/./textmodes /e/emacs/lisp/./erc /e/emacs/lisp/./emacs-lisp /e/emacs/lisp/./calc /e/emacs/lisp/./mh-e /e/emacs/lisp/./calendar /e/emacs/lisp/./language /e/emacs/lisp/./net /e/emacs/lisp/./progmodes /e/emacs/lisp/./gnus /e/emacs/lisp/./international /e/emacs/lisp/./eshell
find /e/emacs/lisp -name "*.elc" -print | xargs chmod +w >/dev/null 2>&1 || true; \
        wd=/e/emacs/lisp; subdirs=`(cd $wd; find . -type d -print)`; for file in $subdirs; do case $file in */Old | */RCS | */CVS | */CVS/* | */.* | */.*/* | */=* ) ;; *) wins="$wins $wd/$file" ;; esac; done; \
        els=`echo $wins | tr ' \011' '\012\012' | \
                sed -e 's|\(.\)$|\1/|' -e 's|^\./||' -e 's|$|*.el|'`;   \
        for el in /e/emacs/lisp/emacs-lisp/byte-opt.el /e/emacs/lisp/emacs-lisp/bytecomp.el /e/emacs/lisp/subr.el /e/emacs/lisp/progmodes/cc-mode.el /e/emacs/lisp/progmodes/cc-vars.el $els; do \
          if test -f $el; \
          then \
            echo Compiling $el; \
            EMACSLOADPATH=/e/emacs/lisp ../src/bootstrap-emacs -batch --no-site-file --multibyte   -f batch-byte-compile-if-not-done $el || exit 1; \
          fi \
        done
Compiling /e/emacs/lisp/emacs-lisp/byte-opt.el
Cannot open load file: encoded-kb
make[2]: *** [compile] Error 1
make[2]: *** Waiting for unfinished jobs....
Cannot open load file: encoded-kb
make[2]: *** [autoloads] Error 255
make[2]: Leaving directory `/e/emacs/lisp'
make[1]: *** [bootstrap-build] Error 2
make[1]: Leaving directory `/e/emacs'
make: *** [bootstrap] Error 2

Since it byte-opt.el requires encoded-kb,
I simply prepended encoded-kb.el to the list:

2007-11-16  Jim Meyering  <jim@meyering.net>

	* Makefile.in (COMPILE_FIRST): Add $(lisp)/international/encoded-kb.el.

diff --git a/lisp/Makefile.in b/lisp/Makefile.in
index efb95b3..746ddef 100644
--- a/lisp/Makefile.in
+++ b/lisp/Makefile.in
@@ -61,6 +61,7 @@ AUTOGENEL = loaddefs.el \
 # bootstrapping.
 
 COMPILE_FIRST = \
+	$(lisp)/international/encoded-kb.el \
 	$(lisp)/emacs-lisp/byte-opt.el \
 	$(lisp)/emacs-lisp/bytecomp.el \
 	$(lisp)/subr.el \
 
------------------------------
BTW, one minor nit: I got this *prompt*.
I would have expected no prompts during a build.

  Loading defgroup from `em-term.el'
  Loading defgroup from `em-unix.el'
  Loading defgroup from `em-xtra.el'
  Delete excess backup versions of /e/emacs/lisp/eshell/esh-groups.el? (y or n)

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Jim Meyering <jim@meyering.net>
> Date: Fri, 16 Nov 2007 12:32:20 +0100
> 
> I've just noticed a bootstrap failure.
> I was building with make -j5:

Does it succeed without -j5?

> Compiling /e/emacs/lisp/emacs-lisp/byte-opt.el
> Cannot open load file: encoded-kb
> make[2]: *** [compile] Error 1
> make[2]: *** Waiting for unfinished jobs....
> Cannot open load file: encoded-kb
> make[2]: *** [autoloads] Error 255
> make[2]: Leaving directory `/e/emacs/lisp'
> make[1]: *** [bootstrap-build] Error 2
> make[1]: Leaving directory `/e/emacs'
> make: *** [bootstrap] Error 2

FWIW, I don't see this problem, neither with nor without -j5, on
GNU/Linux.

> Since it byte-opt.el requires encoded-kb,

It does?  I don't see that dependency on GNU/Linux (and cannot figure
out why there should be such a dependency).  I also tried to
byte-compile byte-opt.el (again, on GNU/Linux), and didn't see any
messages about encoded-kb being loaded.

What platform did you use for this bootstrap?  AFAIK, encoded-kb is
required on MS-Windows, but that's not your case, is it?  And even on
MS-Windows, encoded-kb is loaded during startup, not when compiling
byte-opt.el.

> BTW, one minor nit: I got this *prompt*.
> I would have expected no prompts during a build.
> 
>   Loading defgroup from `em-term.el'
>   Loading defgroup from `em-unix.el'
>   Loading defgroup from `em-xtra.el'
>   Delete excess backup versions of /e/emacs/lisp/eshell/esh-groups.el? (y or n)

I think I fixed this on the trunk, please try again.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

>> Since it byte-opt.el requires encoded-kb,
>
> It does?  I don't see that dependency on GNU/Linux (and cannot figure
> out why there should be such a dependency).

command-line calls set-locale-environment which calls
set-keyboard-coding-system which calls encoded-kbd-setup-display which
autoloads encoded-kb.  You won't notice that unless you have a
nontrivial locale.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Andreas Schwab]

Jim Meyering <jim@meyering.net> writes:

> Compiling /e/emacs/lisp/emacs-lisp/byte-opt.el
> Cannot open load file: encoded-kb

That can only happen when subdirs.el is not up-to-date.  Can you try
again after removing it?

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Cc: Jim Meyering <jim@meyering.net>, emacs-devel@gnu.org
> Date: Sat, 17 Nov 2007 12:27:57 +0100
> 
> You won't notice that unless you have a nontrivial locale.

Which locale loads encoded-kb on GNU/Linux?

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Date: Sat, 17 Nov 2007 12:27:57 +0100
> Cc: Jim Meyering <jim@meyering.net>, emacs-devel@gnu.org
> 
> command-line calls set-locale-environment which calls
> set-keyboard-coding-system which calls encoded-kbd-setup-display which
> autoloads encoded-kb.

Btw, if the above is the correct diagnostics of the problem, then it
should have popped up in the first invocation of bootstrap-emacs,
right?  However, Jim's original report seemed to indicate that Emacs
was already invoked before (although it's easy to be wrong about it,
given that Jim only posted a short fraction of the session
transcript).

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Jim Meyering]

Andreas Schwab <schwab@suse.de> wrote:

> Jim Meyering <jim@meyering.net> writes:
>
>> Compiling /e/emacs/lisp/emacs-lisp/byte-opt.el
>> Cannot open load file: encoded-kb
>
> That can only happen when subdirs.el is not up-to-date.  Can you try
> again after removing it?

Good call.
Then, it worked fine.

BTW, this is with the en_US.UTF-8 locale
on a fedora rawhide system.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Andreas Schwab <schwab@suse.de>
>> Cc: Jim Meyering <jim@meyering.net>, emacs-devel@gnu.org
>> Date: Sat, 17 Nov 2007 12:27:57 +0100
>> 
>> You won't notice that unless you have a nontrivial locale.
>
> Which locale loads encoded-kb on GNU/Linux?

Almost any.  See set-locale-environment.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Date: Sat, 17 Nov 2007 16:23:48 +0100
> Cc: jim@meyering.net, emacs-devel@gnu.org
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> From: Andreas Schwab <schwab@suse.de>
> >> Cc: Jim Meyering <jim@meyering.net>, emacs-devel@gnu.org
> >> Date: Sat, 17 Nov 2007 12:27:57 +0100
> >> 
> >> You won't notice that unless you have a nontrivial locale.
> >
> > Which locale loads encoded-kb on GNU/Linux?
> 
> Almost any.

Then why don't we preload encoded-kb?

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    > Which locale loads encoded-kb on GNU/Linux?

    Almost any.  See set-locale-environment.

But only on ttys, it seems.  For me, Emacs loads encoded-kb
when started on a tty but not when started under X.

I think, therefore, that most Emacs processes will not load
encoded-kb.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Andreas Schwab]

Richard Stallman <rms@gnu.org> writes:

>     > Which locale loads encoded-kb on GNU/Linux?
>
>     Almost any.  See set-locale-environment.
>
> But only on ttys, it seems.  For me, Emacs loads encoded-kb
> when started on a tty but not when started under X.

You are probably not using a non-trivial locale.  Try anything utf-8
based, for example.

> I think, therefore, that most Emacs processes will not load
> encoded-kb.

utf-8 is becoming the norm.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: eliz@gnu.org, jim@meyering.net, emacs-devel@gnu.org
> Date: Sat, 17 Nov 2007 18:31:33 -0500
> 
>     > Which locale loads encoded-kb on GNU/Linux?
> 
>     Almost any.  See set-locale-environment.
> 
> But only on ttys, it seems.

And under -batch, right?

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Richard Stallman <rms@gnu.org>
>> CC: eliz@gnu.org, jim@meyering.net, emacs-devel@gnu.org
>> Date: Sat, 17 Nov 2007 18:31:33 -0500
>> 
>>     > Which locale loads encoded-kb on GNU/Linux?
>> 
>>     Almost any.  See set-locale-environment.
>> 
>> But only on ttys, it seems.
>
> And under -batch, right?

The function set-locale-environment is always called.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Cc: rms@gnu.org, jim@meyering.net, emacs-devel@gnu.org
> Date: Sun, 18 Nov 2007 11:38:16 +0100
> 
> The function set-locale-environment is always called.

Surely, I know this.  The discussion was not about
set-locale-environment, it was about loading encoded-kb.

On many X installations, keyboard input does not need encoded-kb, it
uses X keysyms (as I'm sure you know).

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

> On many X installations, keyboard input does not need encoded-kb, it
> uses X keysyms (as I'm sure you know).

Please read the comment in set-locale-environment.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    >     Almost any.  See set-locale-environment.
    > 
    > But only on ttys, it seems.

    And under -batch, right?

I don't know, but it is an interesting question.  Should Emacs call
set-locale-environment when running in batch mode?

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    > But only on ttys, it seems.  For me, Emacs loads encoded-kb
    > when started on a tty but not when started under X.

    You are probably not using a non-trivial locale.  Try anything utf-8
    based, for example.

What is the name of such a locale?   And how, precisely, should
I specify it?

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Andreas Schwab]

Richard Stallman <rms@gnu.org> writes:

> What is the name of such a locale?   And how, precisely, should
> I specify it?

The charset is the part after the period.  `locale -a' will list all
available locales.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Cc: emacs-devel@gnu.org
> Date: Sun, 18 Nov 2007 20:57:46 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > On many X installations, keyboard input does not need encoded-kb, it
> > uses X keysyms (as I'm sure you know).
> 
> Please read the comment in set-locale-environment.

Really, Andreas, you should try be friendlier with your responses some
day.  It's not a crime to not know something about the latest changes
to Emacs which turn upside down long established memories.

To everybody else who are not in the know: It's the multi-tty change
strikes again: it turns out set-locale-environment was modified to
_always_ set the keyboard encoding:

	  ;; Set the `keyboard-coding-system' if appropriate (tty
	  ;; only).  At least X and MS Windows can generate
	  ;; multilingual input.
	  ;; XXX This was disabled unless `window-system', but that
	  ;; leads to buggy behaviour when a tty frame is opened
	  ;; later.  Setting the keyboard coding system has no adverse
	  ;; effect on X, so let's do it anyway. -- Lorentey
	  (let ((kcs (or coding-system
			 (car (get-language-info language-name
						 'coding-system)))))
	    (if kcs (set-keyboard-coding-system kcs frame)))

So now I'm quite convinced we should preload encoded-kb (unless we
undo this change).

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: schwab@suse.de, jim@meyering.net, emacs-devel@gnu.org
> Date: Sun, 18 Nov 2007 17:45:51 -0500
> 
>     >     Almost any.  See set-locale-environment.
>     > 
>     > But only on ttys, it seems.
> 
>     And under -batch, right?
> 
> I don't know, but it is an interesting question.  Should Emacs call
> set-locale-environment when running in batch mode?

As I wrote elsewhere, it now seems to do that unconditionally.

In general, since -batch may still imply prompts followed by keyboard
input, I think we do need to set up keyboard encoding correctly.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Johan Bockgård]

Eli Zaretskii <eliz@gnu.org> writes:

> 	  ;; Set the `keyboard-coding-system' if appropriate (tty
> 	  ;; only).  At least X and MS Windows can generate
> 	  ;; multilingual input.
> 	  ;; XXX This was disabled unless `window-system', but that

This should rather be "enabled unless" or "disabled when", methinks.

-- 
Johan Bockgård

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Andreas Schwab <schwab@suse.de>
>> Cc: emacs-devel@gnu.org
>> Date: Sun, 18 Nov 2007 20:57:46 +0100
>> 
>> Eli Zaretskii <eliz@gnu.org> writes:
>> 
>> > On many X installations, keyboard input does not need encoded-kb, it
>> > uses X keysyms (as I'm sure you know).
>> 
>> Please read the comment in set-locale-environment.
>
> Really, Andreas, you should try be friendlier with your responses some
> day.

I'm sorry, but I was interrupted and forgot to include the quote.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    > On many X installations, keyboard input does not need encoded-kb, it
    > uses X keysyms (as I'm sure you know).

    Please read the comment in set-locale-environment.

Which comment do you have in mind?  What I see is that
set-keyboard-coding-system is called only if window-system is nil.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

See below the output of `locale' in my configuration.
It looks like I am already using a UTF-8 locale.
But I am using Emacs 22; that's why it doesn't load
encoded-kb under X.

LANG=en_US.UTF-8
LANGUAGE=en
LC_CTYPE="en_US.UTF-8"
LC_NUMERIC="en_US.UTF-8"
LC_TIME="en_US.UTF-8"
LC_COLLATE="en_US.UTF-8"
LC_MONETARY="en_US.UTF-8"
LC_MESSAGES="en_US.UTF-8"
LC_PAPER="en_US.UTF-8"
LC_NAME="en_US.UTF-8"
LC_ADDRESS="en_US.UTF-8"
LC_TELEPHONE="en_US.UTF-8"
LC_MEASUREMENT="en_US.UTF-8"
LC_IDENTIFICATION="en_US.UTF-8"
LC_ALL=

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

	      ;; XXX This was disabled unless `window-system', but that
	      ;; leads to buggy behaviour when a tty frame is opened
	      ;; later.  Setting the keyboard coding system has no adverse
	      ;; effect on X, so let's do it anyway. -- Lorentey

There's one adverse effect: it loads encoded-kb.
We could either preload encoded-kb, or make a change to
do this the first time a tty is opened.

I prefer the latter since most sessions don't use a tty.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    > 
    > I don't know, but it is an interesting question.  Should Emacs call
    > set-locale-environment when running in batch mode?

    As I wrote elsewhere, it now seems to do that unconditionally.

    In general, since -batch may still imply prompts followed by keyboard
    input, I think we do need to set up keyboard encoding correctly.

When Emacs reads input in batch mode, does it do this decoding at all?
I don't know.  But since the input is coming from stdin, not from "the
keyboard", I think it is wrong to do any such decoding.  What if stdin
is a file?

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Stefan Monnier]

>> I don't know, but it is an interesting question.  Should Emacs call
>> set-locale-environment when running in batch mode?

>     As I wrote elsewhere, it now seems to do that unconditionally.

>     In general, since -batch may still imply prompts followed by keyboard
>     input, I think we do need to set up keyboard encoding correctly.

> When Emacs reads input in batch mode, does it do this decoding at all?
> I don't know.  But since the input is coming from stdin, not from "the
> keyboard", I think it is wrong to do any such decoding.

I don't see why it would be wrong.  It would be wrong to do
terminal-escape-sequence processing, but coding-system decoding on the
contrary seems like the right thing to do.


        Stefan

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> Date: Mon, 19 Nov 2007 14:02:37 -0500
> Cc: schwab@suse.de, emacs-devel@gnu.org
> 
> 	      ;; XXX This was disabled unless `window-system', but that
> 	      ;; leads to buggy behaviour when a tty frame is opened
> 	      ;; later.  Setting the keyboard coding system has no adverse
> 	      ;; effect on X, so let's do it anyway. -- Lorentey
> 
> There's one adverse effect: it loads encoded-kb.
> We could either preload encoded-kb, or make a change to
> do this the first time a tty is opened.
> 
> I prefer the latter since most sessions don't use a tty.

On second thought, I also prefer the latter, if someone figures out a
clean way of doing that (I think keyboard input for each tty can in
principle have a different default encoding: e.g., one tty is local,
while the other is remote, and each one belongs to a different
locale).

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: bojohan+news@dd.chalmers.se (Johan =?utf-8?Q?Bockg=C3=A5rd?=)
> Date: Mon, 19 Nov 2007 11:10:12 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > 	  ;; Set the `keyboard-coding-system' if appropriate (tty
> > 	  ;; only).  At least X and MS Windows can generate
> > 	  ;; multilingual input.
> > 	  ;; XXX This was disabled unless `window-system', but that
> 
> This should rather be "enabled unless" or "disabled when", methinks.

Thou shalt never use double negation!

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> Date: Mon, 19 Nov 2007 21:44:14 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: schwab@suse.de, emacs-devel@gnu.org
> 
> > I prefer the latter since most sessions don't use a tty.
> 
> On second thought, I also prefer the latter, if someone figures out a
> clean way of doing that (I think keyboard input for each tty can in
> principle have a different default encoding: e.g., one tty is local,
> while the other is remote, and each one belongs to a different
> locale).

Looking closer at the code, it sounds like we already do TRT:
tty-create-frame-with-faces already calls set-locale-environment.  So
maybe the call in startup.el should be conditioned on the fact that we
are in a GUI session.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Mon, 19 Nov 2007 14:02:38 -0500
> 
>     In general, since -batch may still imply prompts followed by keyboard
>     input, I think we do need to set up keyboard encoding correctly.
> 
> When Emacs reads input in batch mode, does it do this decoding at all?

No, we don't; at least a breakpoint in decode_coding_string does not
break when I call yes-or-no-p in "emacs -batch".  But I think this is
wrong.

> But since the input is coming from stdin, not from "the
> keyboard", I think it is wrong to do any such decoding.  What if stdin
> is a file?

Like Stefan, I think decoding it is TRT under -batch.  It doesn't
matter if it comes from a file: if the user redirects stdin to a file,
she ought to know that the file should be encoded the same as keyboard
input.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Kenichi Handa]

In article <ufxz2arx4.fsf@gnu.org>, Eli Zaretskii <eliz@gnu.org> writes:

> > But since the input is coming from stdin, not from "the
> > keyboard", I think it is wrong to do any such decoding.  What if stdin
> > is a file?

> Like Stefan, I think decoding it is TRT under -batch.

I too agree with that.

---
Kenichi Handa
handa@ni.aist.go.jp

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    Like Stefan, I think decoding it is TRT under -batch.  It doesn't
    matter if it comes from a file: if the user redirects stdin to a file,
    she ought to know that the file should be encoded the same as keyboard
    input.

We could say that, but is it the right thing to say?  It seems to me
that input which isn't coming specifically "from a terminal" should
not get the special treatment for input coming specifically "from a
terminal".

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    > When Emacs reads input in batch mode, does it do this decoding at all?
    > I don't know.  But since the input is coming from stdin, not from "the
    > keyboard", I think it is wrong to do any such decoding.

    I don't see why it would be wrong.  It would be wrong to do
    terminal-escape-sequence processing, but coding-system decoding on the
    contrary seems like the right thing to do.

Could you explain the arguments for this?

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Stefan Monnier]

>> When Emacs reads input in batch mode, does it do this decoding at all?
>> I don't know.  But since the input is coming from stdin, not from "the
>> keyboard", I think it is wrong to do any such decoding.

>     I don't see why it would be wrong.  It would be wrong to do
>     terminal-escape-sequence processing, but coding-system decoding on the
>     contrary seems like the right thing to do.

> Could you explain the arguments for this?

I don't know what to explain: if we assume the input (from stdin) to be
"text" rather than "binary data", then there needs to be
a decoding step.

Maybe you should explain why you think there shouldn't be such a thing?


        Stefan

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Tue, 20 Nov 2007 07:12:25 -0500
> 
>     Like Stefan, I think decoding it is TRT under -batch.  It doesn't
>     matter if it comes from a file: if the user redirects stdin to a file,
>     she ought to know that the file should be encoded the same as keyboard
>     input.
> 
> We could say that, but is it the right thing to say?  It seems to me
> that input which isn't coming specifically "from a terminal" should
> not get the special treatment for input coming specifically "from a
> terminal".

If my keyboard input is in some specific encoding, that generally
means my text files, in particular those I feed instead of keyboard
input, will tend to be encoded the same way.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    I don't know what to explain: if we assume the input (from stdin) to be
    "text" rather than "binary data", then there needs to be
    a decoding step.

    Maybe you should explain why you think there shouldn't be such a thing?

That's not quite what I said.

The input from stdin to a batch program should not be thought of as
keyboard input.  It is just as likely to come from a file or some
other program.

Perhaps it should be decoded somehow, but not the decoding should not
be based on keyboard-coding-system.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Stefan Monnier]

>     I don't know what to explain: if we assume the input (from stdin) to be
>     "text" rather than "binary data", then there needs to be
>     a decoding step.

>     Maybe you should explain why you think there shouldn't be such a thing?

> That's not quite what I said.

> The input from stdin to a batch program should not be thought of as
> keyboard input.  It is just as likely to come from a file or some
> other program.

> Perhaps it should be decoded somehow, but not the decoding should not
> be based on keyboard-coding-system.

Well, some coding-system should be used.  I think the "locale" coding
system makes a lot of sense, and keyboard-coding-system is typically
equal to the locale coding system.

This said, now I'm wondering: in batch mode, does encoded-kb.el have
any effect?  I mean, are input-decode-map and function-key-map used?


        Stefan

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

    If my keyboard input is in some specific encoding, that generally
    means my text files, in particular those I feed instead of keyboard
    input, will tend to be encoded the same way.

All your keyboard input comes from you, but the files on your system
were not all written by you.  That is why Emacs tries to determine
the encoding of a file.  It does not follow keyboard-coding-system.

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Wed, 21 Nov 2007 21:28:12 -0500
> 
>     If my keyboard input is in some specific encoding, that generally
>     means my text files, in particular those I feed instead of keyboard
>     input, will tend to be encoded the same way.
> 
> All your keyboard input comes from you, but the files on your system
> were not all written by you.  That is why Emacs tries to determine
> the encoding of a file.  It does not follow keyboard-coding-system.

If what you say is that we should perhaps try detecting the encoding
of a file input (unlike what we do with keyboard input), then I don't
mind.  The thing is, right now we don't even try to decode input under
"-batch".

Re: parallel bootstrap failure; _prompt_? during bootstrap build

[Richard Stallman]

If someone wants to implement decoding for batch stdin, either based
on guessing the encoding or based on a new global variable, please do.