[PATCH] Disable VC when Emacs is noninteractive -batch

[PATCH] Disable VC when Emacs is noninteractive -batch

[Michael Olson]

[-- Attachment #1.1: Type: text/plain, Size: 713 bytes --]

I've noticed lately that VC can behave badly when trying to build one of
my Emacs Lisp add-on programs.  If a file has been deleted at some
point, it will interrupt the build to prompt me about whether to restore
it -- this is a show-stopper.  Also, even when building Emacs (as of a
week ago, at least), VC prints several spurious messages on the console;
they are harmless in this case, but nonetheless provide needless
clutter.

I think that VC should be disabled by default when emacs is passed the
-batch option.  The attached patch implements this.  Is it OK if I
install it?  VC is supposed to be for working interactively with version
control systems, so I can't see any reason not to install the patch.


[-- Attachment #1.2: Type: application/pgp-signature, Size: 188 bytes --]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: noninteractive-vc.patch --]
[-- Type: text/x-diff, Size: 773 bytes --]

*** lisp/vc-hooks.el	2007-08-05 15:31:30.000000000 -0400
--- lisp/vc-hooks.el.new	2007-08-05 15:30:34.000000000 -0400
***************
*** 62,68 ****
    :type 'regexp
    :group 'vc)
  
! (defcustom vc-handled-backends '(RCS CVS SVN SCCS Bzr Git Hg Arch MCVS)
    ;; Bzr, Git, Hg, Arch and MCVS come last because they are per-tree
    ;; rather than per-dir.
    "List of version control backends for which VC will be used.
--- 62,69 ----
    :type 'regexp
    :group 'vc)
  
! (defcustom vc-handled-backends (if noninteractive nil
!                                  '(RCS CVS SVN SCCS Bzr Git Hg Arch MCVS))
    ;; Bzr, Git, Hg, Arch and MCVS come last because they are per-tree
    ;; rather than per-dir.
    "List of version control backends for which VC will be used.

[-- Attachment #3: Type: text/plain, Size: 255 bytes --]


-- 
       Michael Olson -- FSF Associate Member #652     |
 http://mwolson.org/ -- Jabber: mwolson_at_hcoop.net  |  /` |\ | | |
            Sysadmin -- Hobbies: Lisp, GP2X, HCoop    | |_] | \| |_|
Projects: Emacs, Muse, ERC, EMMS, ErBot, DVC, Planner |

[-- Attachment #4: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Dan Nicolaescu]

Michael Olson <mwolson@gnu.org> writes:

  > I've noticed lately that VC can behave badly when trying to build one of
  > my Emacs Lisp add-on programs.  If a file has been deleted at some
  > point, it will interrupt the build to prompt me about whether to restore
  > it -- this is a show-stopper.  Also, even when building Emacs (as of a
  > week ago, at least), VC prints several spurious messages on the console;
  > they are harmless in this case, but nonetheless provide needless
  > clutter.

Were you working with files under bzr? vc-bzr had a bug that Stefan
fixed recently. 

If not, can you please describe the problem that you are seeing?

  > I think that VC should be disabled by default when emacs is passed the
  > -batch option.  The attached patch implements this.  Is it OK if I
  > install it?  VC is supposed to be for working interactively with version
  > control systems, so I can't see any reason not to install the patch.

If the reason you want to do this to avoid bugs in VC, then IMO this
is not a good idea. If VC has a problem, it has to be fixed.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Michael Olson]

[-- Attachment #1.1: Type: text/plain, Size: 1769 bytes --]

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

> Michael Olson <mwolson@gnu.org> writes:
>
>   > I've noticed lately that VC can behave badly when trying to build
>   > one of my Emacs Lisp add-on programs.  If a file has been deleted
>   > at some point, it will interrupt the build to prompt me about
>   > whether to restore it -- this is a show-stopper.  Also, even when
>   > building Emacs (as of a week ago, at least), VC prints several
>   > spurious messages on the console; they are harmless in this case,
>   > but nonetheless provide needless clutter.
>
> Were you working with files under bzr? vc-bzr had a bug that Stefan
> fixed recently.

Yes, it happened with vc-bzr.

> If not, can you please describe the problem that you are seeing?
>
>   > I think that VC should be disabled by default when emacs is passed
>   > the -batch option.  The attached patch implements this.  Is it OK
>   > if I install it?  VC is supposed to be for working interactively
>   > with version control systems, so I can't see any reason not to
>   > install the patch.
>
> If the reason you want to do this to avoid bugs in VC, then IMO this
> is not a good idea. If VC has a problem, it has to be fixed.

That is not the only reason.  I want vc disabled when in -batch mode
because it is (1) useless there, (2) delays compilation, and (3)
displays many "loading vc-<system>..." messages that clutter the
compilation output.  This bug was just the straw that broke the camel's
back.

-- 
       Michael Olson -- FSF Associate Member #652     |
 http://mwolson.org/ -- Jabber: mwolson_at_hcoop.net  |  /` |\ | | |
            Sysadmin -- Hobbies: Lisp, GP2X, HCoop    | |_] | \| |_|
Projects: Emacs, Muse, ERC, EMMS, ErBot, DVC, Planner |

[-- Attachment #1.2: Type: application/pgp-signature, Size: 188 bytes --]

[-- Attachment #2: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Miles Bader]

Michael Olson <mwolson@gnu.org> writes:
>> If the reason you want to do this to avoid bugs in VC, then IMO this
>> is not a good idea. If VC has a problem, it has to be fixed.
>
> That is not the only reason.  I want vc disabled when in -batch mode
> because it is (1) useless there, (2) delays compilation, and (3)
> displays many "loading vc-<system>..." messages that clutter the
> compilation output.

Yeah, I agree with Michael, VC shouldn't be involved in batch mode.

There are plenty of other opportunities to find bugs in the vc code.

-Miles

-- 
97% of everything is grunge

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Stephen J. Turnbull]

Dan Nicolaescu writes:

 > If the reason you want to do this to avoid bugs in VC, then IMO this
 > is not a good idea. If VC has a problem, it has to be fixed.

IMO, if VC's hooks are active in batch mode, that's a bug.  You don't
want programs that are *implicitly* sensitive to environment state
active in batch mode.  (I'm sure there are good reasons for using VC
noninteractively, although I personally don't use it in batch mode.
But such programs should call VC functions explicitly.)

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    IMO, if VC's hooks are active in batch mode, that's a bug.

I don't see why.

      You don't
    want programs that are *implicitly* sensitive to environment state
    active in batch mode.

I don't follow at all.  If "environment state" means the file system,
we certainly don't expect commands that operate on files to be
insensitive to that.

      (I'm sure there are good reasons for using VC
    noninteractively, although I personally don't use it in batch mode.
    But such programs should call VC functions explicitly.)

VC only does anything when you call VC functions explicitly.
Unless it has a bug.

    > That is not the only reason.  I want vc disabled when in -batch mode
    > because it is (1) useless there, (2) delays compilation, and (3)
    > displays many "loading vc-<system>..." messages that clutter the
    > compilation output.

It would be harmless and easy to turn off VC for batch mode
compilation.  We just have to change the commands that do this
compilation:

*** bytecomp.el	26 Jul 2007 12:57:26 -0400	2.204
--- bytecomp.el	06 Aug 2007 11:47:18 -0400	
***************
*** 4124,4130 ****
    (defvar command-line-args-left)	;Avoid 'free variable' warning
    (if (not noninteractive)
        (error "`batch-byte-compile' is to be used only with -batch"))
!   (let ((error nil))
      (while command-line-args-left
        (if (file-directory-p (expand-file-name (car command-line-args-left)))
  	  ;; Directory as argument.
--- 4124,4131 ----
    (defvar command-line-args-left)	;Avoid 'free variable' warning
    (if (not noninteractive)
        (error "`batch-byte-compile' is to be used only with -batch"))
!   (let ((error nil)
! 	(vc-handled-backends nil))
      (while command-line-args-left
        (if (file-directory-p (expand-file-name (car command-line-args-left)))
  	  ;; Directory as argument.
***************
*** 4192,4198 ****
    (or command-line-args-left
        (setq command-line-args-left '(".")))
    (while command-line-args-left
!     (byte-recompile-directory (car command-line-args-left) arg)
      (setq command-line-args-left (cdr command-line-args-left)))
    (kill-emacs 0))
  
--- 4193,4200 ----
    (or command-line-args-left
        (setq command-line-args-left '(".")))
    (while command-line-args-left
!     (let ((vc-handled-backends nil))
!       (byte-recompile-directory (car command-line-args-left) arg))
      (setq command-line-args-left (cdr command-line-args-left)))
    (kill-emacs 0))

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Stephen J. Turnbull]

Richard Stallman writes:

  stephen wrote:

 >     IMO, if VC's hooks are active in batch mode, that's a bug.
 > 
 > I don't see why.

Because they change the behavior of Emacs depending on, for example,
whether the working file differs from the repository file.

 >     You don't want programs that are *implicitly* sensitive to
 >     environment state active in batch mode.
 > 
 > I don't follow at all.  If "environment state" means the file system,
 > we certainly don't expect commands that operate on files to be
 > insensitive to that.

VC, however, uses metadata that most programs are not aware of.  This
can make a program that is deterministic when VC is not loaded
nondeterministic when it is loaded.  That's a bad thing if no human is
there to say "Whoa!  M-x undo!"

 >       (I'm sure there are good reasons for using VC
 >     noninteractively, although I personally don't use it in batch mode.
 >     But such programs should call VC functions explicitly.)
 > 
 > VC only does anything when you call VC functions explicitly.
 > Unless it has a bug.

That's clearly not true, for some value of "does anything": VC
installs hooks which are called for various file and buffer
operations.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

     >     IMO, if VC's hooks are active in batch mode, that's a bug.
     > 
     > I don't see why.

    Because they change the behavior of Emacs depending on, for example,
    whether the working file differs from the repository file.

Only if you use VC commands, as far as I know.

    VC, however, uses metadata that most programs are not aware of.  This
    can make a program that is deterministic when VC is not loaded
    nondeterministic when it is loaded.

Can you explain concretely what you mean?

     > VC only does anything when you call VC functions explicitly.
     > Unless it has a bug.

    That's clearly not true, for some value of "does anything": VC
    installs hooks which are called for various file and buffer
    operations.

I don't follow you.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Stephen J. Turnbull]

Richard Stallman writes:

 >     Because they change the behavior of Emacs depending on, for example,
 >     whether the working file differs from the repository file.
 > 
 > Only if you use VC commands, as far as I know.

I'm sorry, you're right.

A friend had a problem a couple years back but on review, I see the
script was loading her init file, which require'd vc-hooks.  So VC
hooks are not going to affect carefully written batch programs, as far
as I can see.

 >      > VC only does anything when you call VC functions explicitly.
 >      > Unless it has a bug.
 > 
 >     That's clearly not true, for some value of "does anything": VC
 >     installs hooks which are called for various file and buffer
 >     operations.

I'm thinking specifically of using `find-file' on what you think is a
non-existent file, and discovering that the buffer has non-empty
content because the file has been checked out of RCS.  I would not
call that "explicitly calling VC functions".

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    I'm thinking specifically of using `find-file' on what you think is a
    non-existent file, and discovering that the buffer has non-empty
    content because the file has been checked out of RCS.  I would not
    call that "explicitly calling VC functions".

That is true.  However, in that case, given the way RCS is used, I
think that checking out the file is the right thing to do.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Michael Olson]

[-- Attachment #1.1: Type: text/plain, Size: 1350 bytes --]

Richard Stallman <rms@gnu.org> writes:

>       (I'm sure there are good reasons for using VC
>     noninteractively, although I personally don't use it in batch mode.
>     But such programs should call VC functions explicitly.)
>
> VC only does anything when you call VC functions explicitly.
> Unless it has a bug.

Please name even one good use for VC in batch/noninteractive mode.  I
can't think of a single one, myself, and knowing a good use would allow
me to adjust my argument accordingly.

>     > That is not the only reason.  I want vc disabled when in -batch mode
>     > because it is (1) useless there, (2) delays compilation, and (3)
>     > displays many "loading vc-<system>..." messages that clutter the
>     > compilation output.
>
> It would be harmless and easy to turn off VC for batch mode
> compilation.  We just have to change the commands that do this
> compilation:

This is much less simple than my patch.  And it doesn't cover the case
for generating autoloads, or running elint in batch mode via "make test"
for one of my projects.

-- 
       Michael Olson -- FSF Associate Member #652     |
 http://mwolson.org/ -- Jabber: mwolson_at_hcoop.net  |  /` |\ | | |
            Sysadmin -- Hobbies: Lisp, GP2X, HCoop    | |_] | \| |_|
Projects: Emacs, Muse, ERC, EMMS, ErBot, DVC, Planner |

[-- Attachment #1.2: Type: application/pgp-signature, Size: 188 bytes --]

[-- Attachment #2: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Jason Rumney]

Michael Olson wrote:
> Please name even one good use for VC in batch/noninteractive mode.  I
> can't think of a single one, myself, and knowing a good use would allow
> me to adjust my argument accordingly.
>   

It's not too far fetched to imagine someone writing a Lisp program to
perform some VC functions and running it in batch mode.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    Please name even one good use for VC in batch/noninteractive mode.  I
    can't think of a single one, myself, and knowing a good use would allow
    me to adjust my argument accordingly.

Relicensing a program is the sort of situation where you might
want to do this.

We might want to arrange to automatically answer "yes" when it
asks whether to check out a deleted a file, if in batch mode.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[David Kastrup]

Richard Stallman <rms@gnu.org> writes:

>     Please name even one good use for VC in batch/noninteractive mode.  I
>     can't think of a single one, myself, and knowing a good use would allow
>     me to adjust my argument accordingly.
>
> Relicensing a program is the sort of situation where you might
> want to do this.
>
> We might want to arrange to automatically answer "yes" when it
> asks whether to check out a deleted a file, if in batch mode.

That's plain creepy as far as I am concerned.  I think that we should
have the following behavior for RCS:

If you open a non-existing file under RCS control, you get an empty
buffer, but with RCS feeling responsible for it.  C-x v v checks out
the file after asking whether you also want to lock it.

In that way, Emacs won't change its behavior due to VC unless one
actually calls VC commands.  And it also won't invoke any executables
unasked for.

-- 
David Kastrup

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    > We might want to arrange to automatically answer "yes" when it
    > asks whether to check out a deleted a file, if in batch mode.

    That's plain creepy as far as I am concerned.

Apparently the Make maintainers didn't consider it "creepy".

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[David Kastrup]

Richard Stallman <rms@gnu.org> writes:

>     > We might want to arrange to automatically answer "yes" when it
>     > asks whether to check out a deleted a file, if in batch mode.
>
>     That's plain creepy as far as I am concerned.
>
> Apparently the Make maintainers didn't consider it "creepy".

Please, that's absurd.  It is the _primary_ _purpose_ of Make to
create missing files from existing ones, if not otherwise possible, by
implied rules.

But it is not the purpose of Emacs to magically generate files when I
open them.  It does not run makeinfo when I try opening a non-existing
info file.  And it certainly should not do something like that in a
batch job.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Michael Olson]

[-- Attachment #1.1: Type: text/plain, Size: 1971 bytes --]

Richard Stallman <rms@gnu.org> writes:

>     Please name even one good use for VC in batch/noninteractive mode.
>     I can't think of a single one, myself, and knowing a good use
>     would allow me to adjust my argument accordingly.
>
> Relicensing a program is the sort of situation where you might
> want to do this.

What part does VC play in such a change?

> We might want to arrange to automatically answer "yes" when it
> asks whether to check out a deleted a file, if in batch mode.

For any backend other than RCS, that would be bad.  When I delete a
file, I don't want it to come back during compilation.  That could lead
to confusing scenarios where something works in development versions of
a program (because a bug is hidden by the presence of the deleted file),
but fails to work for a released version.  Further, it would be just
plain annoying, because I would have to remember to re-delete the file
before checking in any change.

If we did insist on keeping VC active in batch mode (which I still don't
see a good reason to do), then for any backend other than RCS, the
default answer to that question ought to be "no" rather than "yes".
Further, even when using RCS, maintainers ought not to depend on Emacs
to check out a missing file for them -- they should instead have a
Makefile target which ensures that the file is checked out properly
before compiling it or doing other work with it.  Otherwise they would
be forcing people to use Emacs when developing their projects, which
would create resentment.  (Admittedly, this is perhaps a contrived case,
since I don't know of any collaborative projects that still use RCS, but
the principle remains.)

-- 
       Michael Olson -- FSF Associate Member #652     |
 http://mwolson.org/ -- Jabber: mwolson_at_hcoop.net  |  /` |\ | | |
            Sysadmin -- Hobbies: Lisp, GP2X, HCoop    | |_] | \| |_|
Projects: Emacs, Muse, ERC, EMMS, ErBot, DVC, Planner |

[-- Attachment #1.2: Type: application/pgp-signature, Size: 188 bytes --]

[-- Attachment #2: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    > Relicensing a program is the sort of situation where you might
    > want to do this.

    What part does VC play in such a change?

You would use it to check in the files after relicensing.

    > We might want to arrange to automatically answer "yes" when it
    > asks whether to check out a deleted a file, if in batch mode.

    For any backend other than RCS, that would be bad.

This feature is meant for RCS.  Someone else proposed to turn
it off completely for other back ends.  That seems good to me.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Stefan Monnier]

>> Relicensing a program is the sort of situation where you might
>> want to do this.

>     What part does VC play in such a change?

> You would use it to check in the files after relicensing.

Most likely, you'd want a single commit for all the files, so except in
backends like RCS and SCCS which only work on a file-by-file basis, the
current VC wouldn't be of any use (when it gets fileset support that may
change somewhat).

> This feature is meant for RCS.  Someone else proposed to turn
> it off completely for other back ends.

That was me.

> That seems good to me.

OK, I'll make the change.


        Stefan

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Michael Olson]

[-- Attachment #1.1: Type: text/plain, Size: 730 bytes --]

Richard Stallman <rms@gnu.org> writes:

>     > Relicensing a program is the sort of situation where you might
>     > want to do this.
>
>     What part does VC play in such a change?
>
> You would use it to check in the files after relicensing.

I would think that you'd want to review the changes manually before
committing such a large change.  Committing automatically (which is what
I presume that you would use VC for, in this case) seems dangerous.

-- 
       Michael Olson -- FSF Associate Member #652     |
 http://mwolson.org/ -- Jabber: mwolson_at_hcoop.net  |  /` |\ | | |
            Sysadmin -- Hobbies: Lisp, GP2X, HCoop    | |_] | \| |_|
Projects: Emacs, Muse, ERC, EMMS, ErBot, DVC, Planner |

[-- Attachment #1.2: Type: application/pgp-signature, Size: 188 bytes --]

[-- Attachment #2: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    I would think that you'd want to review the changes manually before
    committing such a large change.  Committing automatically (which is what
    I presume that you would use VC for, in this case) seems dangerous.

You might do either one, depending.  The point is that this is
a reasonable use for VC commands in batch mode.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Edward O'Connor]

Michael Olson wrote:

> This is much less simple than my patch.  And it doesn't cover the case
> for generating autoloads, or running elint in batch mode via "make test"
> for one of my projects.

Speaking of which, would you consider reformulating
`planner-elint-files' into a patch for elint itself? Perhaps two new
functions, `elint-file' (a command) and `batch-elint-files'. I almost
never remember to use elint, and by default it's only available for
interactive use. If I could add such a Makefile rule, I might remember
to use it more often.


Ted

-- 
Edward O'Connor
hober0@gmail.com

Ense petit placidam sub libertate quietem.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    I think that VC should be disabled by default when emacs is passed the
    -batch option.

Please don't do it.  There is no logical reason why it is wrong for
batch-mode Emacs to do VC operations, so I think that is an attempt to
work around a problem.  We should first try to solve it the right way.

      Also, even when building Emacs (as of a
    week ago, at least), VC prints several spurious messages on the console;
    they are harmless in this case, but nonetheless provide needless
    clutter.

What causes these messages?

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Stefan Monnier]

> my Emacs Lisp add-on programs.  If a file has been deleted at some
> point, it will interrupt the build to prompt me about whether to restore
> it -- this is a show-stopper.  Also, even when building Emacs (as of a

Actually I think this is a "feature" of VC which we should tame
significantly: basically, this only makes sense under RCS (and maybe SCCS)
but happens to be in the generic part of VC, so it can bite at any time when
used with other backends.  The idea was that under RCS if a file <foo> is
absent but RCS/<foo>,v is present, than <foo> is virtually present, except
it needs to be checkout first.  With current systems, the checkout is done
once and for all files, outside of VC, so the feature makes little
sense there.

I suggest to rename vc-default-find-file-not-found-hook into
vc-rcs-find-file-not-found-hook and create a new
vc-default-find-file-not-found-hook which is just an alias for `ignore'.

vc-arch.el is already careful to define a vc-arch-find-file-not-found-hook
which overrides vc-default-find-file-not-found-hook (and does nothing) and
I think most/all other systems would benefit from such a change.


        Stefan

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[David Kastrup]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> my Emacs Lisp add-on programs.  If a file has been deleted at some
>> point, it will interrupt the build to prompt me about whether to restore
>> it -- this is a show-stopper.  Also, even when building Emacs (as of a
>
> Actually I think this is a "feature" of VC which we should tame
> significantly: basically, this only makes sense under RCS (and maybe SCCS)
> but happens to be in the generic part of VC, so it can bite at any time when
> used with other backends.  The idea was that under RCS if a file <foo> is
> absent but RCS/<foo>,v is present, than <foo> is virtually present, except
> it needs to be checkout first.

That does not even make sense in RCS for me.  One locks files, but
does not delete them.

> vc-arch.el is already careful to define a
> vc-arch-find-file-not-found-hook which overrides
> vc-default-find-file-not-found-hook (and does nothing) and I think
> most/all other systems would benefit from such a change.

_Including_ RCS.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Miles Bader]

David Kastrup <dak@gnu.org> writes:
> That does not even make sense in RCS for me.  One locks files, but
> does not delete them.

Perhaps it doesn't make a lot of sense, but IIRC, it was a non-uncommon
practice a rather long time ago.  Note that "make" has similar
auto-checkout functionality to handle this case...  [well I don't know
about current versions of GNU make, but it used to, anyway.]

-Miles

-- 
自らを空にして、心を開く時、道は開かれる

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Eli Zaretskii]

> From: David Kastrup <dak@gnu.org>
> Date: Mon, 06 Aug 2007 21:16:44 +0200
> Cc: Michael Olson <mwolson@gnu.org>, emacs-devel@gnu.org
> 
> > Actually I think this is a "feature" of VC which we should tame
> > significantly: basically, this only makes sense under RCS (and maybe SCCS)
> > but happens to be in the generic part of VC, so it can bite at any time when
> > used with other backends.  The idea was that under RCS if a file <foo> is
> > absent but RCS/<foo>,v is present, than <foo> is virtually present, except
> > it needs to be checkout first.
> 
> That does not even make sense in RCS for me.  One locks files, but
> does not delete them.

AFAIR, with RCS, when you checkin a file, the work copy is deleted
unless you use the -u or -l switch.  So it's quite normal to have no
working file at all.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[David Kastrup]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: David Kastrup <dak@gnu.org>
>> Date: Mon, 06 Aug 2007 21:16:44 +0200
>> Cc: Michael Olson <mwolson@gnu.org>, emacs-devel@gnu.org
>> 
>> > Actually I think this is a "feature" of VC which we should tame
>> > significantly: basically, this only makes sense under RCS (and maybe SCCS)
>> > but happens to be in the generic part of VC, so it can bite at any time when
>> > used with other backends.  The idea was that under RCS if a file <foo> is
>> > absent but RCS/<foo>,v is present, than <foo> is virtually present, except
>> > it needs to be checkout first.
>> 
>> That does not even make sense in RCS for me.  One locks files, but
>> does not delete them.
>
> AFAIR, with RCS, when you checkin a file, the work copy is deleted
> unless you use the -u or -l switch.  So it's quite normal to have no
> working file at all.

Indeed, ci(1).  Wow, that is really sick.  Anyway, the normal behavior
for Emacs is to keep locked files around, so I'd say we should cater
for that case primordinarily.  If that means some slight inconvenience
in "classical" RCS use from the command line, presumably not getting
notice of the virtual file presence until checked out, it should be
something we can live with.

I don't think that many people use RCS from command line and from
within Emacs interspersed.

-- 
David Kastrup

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Michael Olson]

[-- Attachment #1.1: Type: text/plain, Size: 1131 bytes --]

David Kastrup <dak@gnu.org> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>> AFAIR, with RCS, when you checkin a file, the work copy is deleted
>> unless you use the -u or -l switch.  So it's quite normal to have no
>> working file at all.
>
> Indeed, ci(1).  Wow, that is really sick.  Anyway, the normal behavior
> for Emacs is to keep locked files around, so I'd say we should cater
> for that case primordinarily.  If that means some slight inconvenience
> in "classical" RCS use from the command line, presumably not getting
> notice of the virtual file presence until checked out, it should be
> something we can live with.
>
> I don't think that many people use RCS from command line and from
> within Emacs interspersed.

In this case, I wouldn't mind if just '(rcs) was bound to
vc-handled-backends when noninteractive, rather than nil.

-- 
       Michael Olson -- FSF Associate Member #652     |
 http://mwolson.org/ -- Jabber: mwolson_at_hcoop.net  |  /` |\ | | |
            Sysadmin -- Hobbies: Lisp, GP2X, HCoop    | |_] | \| |_|
Projects: Emacs, Muse, ERC, EMMS, ErBot, DVC, Planner |

[-- Attachment #1.2: Type: application/pgp-signature, Size: 188 bytes --]

[-- Attachment #2: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    Indeed, ci(1).  Wow, that is really sick.  Anyway, the normal behavior
    for Emacs is to keep locked files around, so I'd say we should cater
    for that case primordinarily.  If that means some slight inconvenience
    in "classical" RCS use from the command line, presumably not getting
    notice of the virtual file presence until checked out, it should be
    something we can live with.

It would be better, and nearly as easy, to default the answer to "yes"
when Emacs in batch mode asks about recreating such a file.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[David Kastrup]

Richard Stallman <rms@gnu.org> writes:

>     Indeed, ci(1).  Wow, that is really sick.  Anyway, the normal behavior
>     for Emacs is to keep locked files around, so I'd say we should cater
>     for that case primordinarily.  If that means some slight inconvenience
>     in "classical" RCS use from the command line, presumably not getting
>     notice of the virtual file presence until checked out, it should be
>     something we can live with.
>
> It would be better, and nearly as easy, to default the answer to "yes"
> when Emacs in batch mode asks about recreating such a file.

Sorry, but I really don't think Emacs should create non-existing files
without being asked for it.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    Sorry, but I really don't think Emacs should create non-existing files
    without being asked for it.

The point is, with RCS, these files are not really nonexistent.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[David Kastrup]

Richard Stallman <rms@gnu.org> writes:

>     Sorry, but I really don't think Emacs should create non-existing files
>     without being asked for it.
>
> The point is, with RCS, these files are not really nonexistent.

We have to agree to disagree on that point.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[chad brown]

Anecdotal evidence here, but:

These days, my primary use of RCS is to keep copies of files (with  
history) that I no longer want in my working directories.  I  
occasionally examine these files, but it's pretty rare - basically a  
warm archive, with history.

I would be leery of having emacs automatically answer any VC  
questions at all -- there are new vc systems being added and  
developed, and it's not hard for me to imagine a (non-emacs-using)  
developer adding a prompt where either `yes' or `no' is the right  
answer most of the time but a destructively wrong answer in some  
rarer cases.

*chad

On Aug 8, 2007, at 5:17 PM, David Kastrup wrote:

> Richard Stallman <rms@gnu.org> writes:
>
>>     Sorry, but I really don't think Emacs should create non- 
>> existing files
>>     without being asked for it.
>>
>> The point is, with RCS, these files are not really nonexistent.
>
> We have to agree to disagree on that point.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    These days, my primary use of RCS is to keep copies of files (with  
    history) that I no longer want in my working directories.  I  
    occasionally examine these files, but it's pretty rare - basically a  
    warm archive, with history.

Do you ever delete the source files that are stored in RCS?
If not, then this feature will never do anything in your case.
So why object?

    I would be leery of having emacs automatically answer any VC  
    questions at all -- there are new vc systems being added and  
    developed, and it's not hard for me to imagine a (non-emacs-using)  
    developer adding a prompt where either `yes' or `no' is the right  
    answer most of the time but a destructively wrong answer in some  
    rarer cases.

Speculations about hypothetical future features, that will probably
never exist and would be separate issues if they did exist, does not
seem relevant.

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[chad brown]

On Aug 12, 2007, at 10:56 AM, Richard Stallman wrote:
>
> Do you ever delete the source files that are stored in RCS?
> If not, then this feature will never do anything in your case.
> So why object?

I'm not sure that I understand your question, but maybe I can clarify  
via example:  ~/Library/elisp/RCS/mh-flists.el,v  exists and contains  
valuable (to me) historic information about changes over time, but ~/ 
Library/elisp/mh-flists.el does not exist,  should not exist, and has  
not existed for many months.

I had read the discussion to suggest that emacs would, under certain  
circumstances, silently recreate the later file from the former.  If  
I am mistaken, please forgive my interruption -- I have been moving  
recently, and might easily have missed some detail while reading  
emacs-devel in bits and spurts.

Regardless, I merely hoped to aid by describing some (possibly  
unusual) usage patterns that I have held for years; I'm sure that I  
can adapt myself and/or emacs for my own usage.

On Aug 14, 2007, at 8:46 AM, Davis Herring wrote:

> It's not the VC system -- that is, the external program like "cvs"  
> -- that
> asks such questions, but rather the VC package in Emacs.  There's  
> no risk
> that people working on the latter won't be Emacs users and won't  
> (if it's
> documented properly!) understand the implications of automatic  
> answers.

Ah, I see that I misunderstood this part.  Mea Culpa, and thanks for  
setting me straight.

*chad

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Richard Stallman]

    I'm not sure that I understand your question, but maybe I can clarify  
    via example:  ~/Library/elisp/RCS/mh-flists.el,v  exists and contains  
    valuable (to me) historic information about changes over time, but ~/ 
    Library/elisp/mh-flists.el does not exist,  should not exist, and has  
    not existed for many months.

What is the motive for doing things this way?
Is it considered normal RCS usage?

Why not rename the master file
to some name that nothing will try to use,
such as ~/Library/elisp/RCS/=mh-flists.el.old,v
or ~/Library/.archive/elisp/RCS/mh-flists.el,v
?

Re: [PATCH] Disable VC when Emacs is noninteractive -batch

[Davis Herring]

> I would be leery of having emacs automatically answer any VC
> questions at all -- there are new vc systems being added and
> developed, and it's not hard for me to imagine a (non-emacs-using)
> developer adding a prompt where either `yes' or `no' is the right
> answer most of the time but a destructively wrong answer in some
> rarer cases.

It's not the VC system -- that is, the external program like "cvs" -- that
asks such questions, but rather the VC package in Emacs.  There's no risk
that people working on the latter won't be Emacs users and won't (if it's
documented properly!) understand the implications of automatic answers.

Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or
too sparse, it is because mass-energy conversion has occurred during
shipping.