case's definition is void when replying to mail

case's definition is void when replying to mail

[Raphael Kubo da Costa]

Hey there,

After commit 4d989747fbb1b5011f15c9ddc1d642fa0681c08e, I usually get the
following message when replying to mail (S W or S V):

  Symbol's function definition is void: case

debug-on-error tells me that the problem is in the `case` call in message-yank-
original, which is only defined after 'cl is required.

This is GNU Emacs 23.2 -- in another machine with GNU Emacs 23.3, starting it
and typing 'C-h f case' showed me the right definition right away.

Re: case's definition is void when replying to mail

[Ted Zlatanov]

On Fri, 11 Mar 2011 13:47:18 +0000 (UTC) Raphael Kubo da Costa <kubito@gmail.com> wrote: 

RKdC> Hey there,
RKdC> After commit 4d989747fbb1b5011f15c9ddc1d642fa0681c08e, I usually get the
RKdC> following message when replying to mail (S W or S V):

RKdC>   Symbol's function definition is void: case

RKdC> debug-on-error tells me that the problem is in the `case` call in message-yank-
RKdC> original, which is only defined after 'cl is required.

RKdC> This is GNU Emacs 23.2 -- in another machine with GNU Emacs 23.3, starting it
RKdC> and typing 'C-h f case' showed me the right definition right away.

I believe this is fixed in the Gnus trunk (also reported by David Kastrup)..

Ted

Re: case's definition is void when replying to mail

[Raphael Kubo da Costa]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Fri, 11 Mar 2011 13:47:18 +0000 (UTC) Raphael Kubo da Costa <kubito@gmail.com> wrote: 
>
> RKdC> Hey there,
> RKdC> After commit 4d989747fbb1b5011f15c9ddc1d642fa0681c08e, I usually get the
> RKdC> following message when replying to mail (S W or S V):
>
> RKdC>   Symbol's function definition is void: case
>
> RKdC> debug-on-error tells me that the problem is in the `case` call in message-yank-
> RKdC> original, which is only defined after 'cl is required.
>
> RKdC> This is GNU Emacs 23.2 -- in another machine with GNU Emacs 23.3, starting it
> RKdC> and typing 'C-h f case' showed me the right definition right away.
>
> I believe this is fixed in the Gnus trunk (also reported by David
> Kastrup)..

Yep, your commit ab3426f did the trick, thanks :)

Re: case's definition is void when replying to mail

[Raphael Kubo da Costa]

Raphael Kubo da Costa <kubito <at> gmail.com> writes:

> Yep, your commit ab3426f did the trick, thanks :)

Commit 5edabb7 kind of reverted your commit ab3426f and reintroduced the bug.

Re: case's definition is void when replying to mail

[Ted Zlatanov]

On Wed, 16 Mar 2011 16:26:46 +0000 (UTC) Raphael Kubo da Costa <kubito@gmail.com> wrote: 

RKdC> Raphael Kubo da Costa <kubito <at> gmail.com> writes:
>> Yep, your commit ab3426f did the trick, thanks :)

RKdC> Commit 5edabb7 kind of reverted your commit ab3426f and reintroduced the bug.

Stefan Monnier preferred `case' there and I don't care much.

So you're saying that

(eval-when-compile
  (require 'cl))

doesn't work for you (it's at the top of message.el).  Isn't that
supposed to work?  Sorry but I don't know eval-* very well.

Ted

Re: case's definition is void when replying to mail

[Raphael Kubo da Costa]

Ted Zlatanov <tzz <at> lifelogs.com> writes:

> So you're saying that
> 
> (eval-when-compile
>   (require 'cl))
> 
> doesn't work for you (it's at the top of message.el).  Isn't that
> supposed to work?  Sorry but I don't know eval-* very well.

I don't really know what's going on (and I only experience the issue in a single
machine, which now has GNU Emacs 23.3) -- I don't see any errors or warnings
related to this when compiling, though.

Any ideas how to help debug this?

Re: case's definition is void when replying to mail

[Ted Zlatanov]

On Wed, 16 Mar 2011 17:34:30 +0000 (UTC) Raphael Kubo da Costa <kubito@gmail.com> wrote: 

RKdC> Ted Zlatanov <tzz <at> lifelogs.com> writes:
>> So you're saying that
>> 
>> (eval-when-compile
>> (require 'cl))
>> 
>> doesn't work for you (it's at the top of message.el).  Isn't that
>> supposed to work?  Sorry but I don't know eval-* very well.

RKdC> I don't really know what's going on (and I only experience the issue in a single
RKdC> machine, which now has GNU Emacs 23.3) -- I don't see any errors or warnings
RKdC> related to this when compiling, though.

RKdC> Any ideas how to help debug this?

Try (require 'cl) in addition to the `eval-when-compile'.  Does it fix it?

When you remove the `eval-when-compile', does it work?  Do compiles
work?

Ted

Re: case's definition is void when replying to mail

[Raphael Kubo da Costa]

Ted Zlatanov <tzz <at> lifelogs.com> writes:

> On Wed, 16 Mar 2011 17:34:30 +0000 (UTC) Raphael Kubo da Costa <kubito <at>
gmail.com> wrote: 
> 
> RKdC> Any ideas how to help debug this?
> 
> Try (require 'cl) in addition to the `eval-when-compile'.  Does it fix it?
> 
> When you remove the `eval-when-compile', does it work?  Do compiles
> work?

Yes, if I directly (require 'cl) or use `eval-and-compile' instead of `eval-
when-compile', everything works.

I can see that many other files also use the `eval-when-compile+case' dance
without (apparently) breaking anything (I haven't looked close enough to know if
I have ever reached any of the code that uses it).

For some reason, that `case' macro is being interpreted at run-time instead of
compile-time. It works in the other machines where I use Gnus because I have
`color-theme-initialize' in my ~/.emacs, which ends up calling `(require 'cl)'
on its own.

Re: case's definition is void when replying to mail

[Antoine Levitt]

17/03/11 14:50, Raphael Kubo da Costa
> Ted Zlatanov <tzz <at> lifelogs.com> writes:
>
>> On Wed, 16 Mar 2011 17:34:30 +0000 (UTC) Raphael Kubo da Costa <kubito <at>
> gmail.com> wrote: 
>> 
>> RKdC> Any ideas how to help debug this?
>> 
>> Try (require 'cl) in addition to the `eval-when-compile'.  Does it fix it?
>> 
>> When you remove the `eval-when-compile', does it work?  Do compiles
>> work?
>
> Yes, if I directly (require 'cl) or use `eval-and-compile' instead of `eval-
> when-compile', everything works.
>
> I can see that many other files also use the `eval-when-compile+case' dance
> without (apparently) breaking anything (I haven't looked close enough to know if
> I have ever reached any of the code that uses it).
>
> For some reason, that `case' macro is being interpreted at run-time instead of
> compile-time. It works in the other machines where I use Gnus because I have
> `color-theme-initialize' in my ~/.emacs, which ends up calling `(require 'cl)'
> on its own.

Maybe something's wrong in your build chain? Try byte compiling the file
again, or if that doesn't work, do a make clean && make bootstrap.

Re: case's definition is void when replying to mail

[Raphael Kubo da Costa]

Antoine Levitt <antoine.levitt <at> gmail.com> writes:

> Maybe something's wrong in your build chain? Try byte compiling the file
> again, or if that doesn't work, do a make clean && make bootstrap.

I always `make clean && ./configure && make' after fetching and merging from
git. And I couldn't find a `bootstrap' target for make.

Re: case's definition is void when replying to mail

[Lars Magne Ingebrigtsen]

Raphael Kubo da Costa <kubito@gmail.com> writes:

> Commit 5edabb7 kind of reverted your commit ab3426f and reintroduced
> the bug.

`(setq debug-on-error t)', repeat the bug and post the resulting
backtrace.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: case's definition is void when replying to mail

[Raphael Kubo da Costa]

Lars Magne Ingebrigtsen wrote:

> Raphael Kubo da Costa <kubito@gmail.com> writes:
> 
>> Commit 5edabb7 kind of reverted your commit ab3426f and reintroduced
>> the bug.
> 
> `(setq debug-on-error t)', repeat the bug and post the resulting
> backtrace.

When using "S V" on a random IMAP message:

Debugger entered--Lisp error: (void-function case)
  (case message-cite-reply-position (above (message-goto-body) (insert body-
text) (insert ...) (message-goto-body)) (below (message-goto-signature)))
  (progn (when (equal message-cite-reply-position ...) (save-excursion ... 
...)) (if (bufferp message-reply-buffer) (delete-windows-on message-reply-
buffer t)) (push-mark (save-excursion ... ... ...)) (unless arg (funcall 
message-cite-function) (unless ... ...)) (case message-cite-reply-position 
(above ... ... ... ...) (below ...)) (unless modified (setq message-checksum 
...)))
  (if (and message-reply-buffer message-cite-function) (progn (when ... ...) 
(if ... ...) (push-mark ...) (unless arg ... ...) (case message-cite-reply-
position ... ...) (unless modified ...)))
  (when (and message-reply-buffer message-cite-function) (when (equal 
message-cite-reply-position ...) (save-excursion ... ...)) (if (bufferp 
message-reply-buffer) (delete-windows-on message-reply-buffer t)) (push-mark 
(save-excursion ... ... ...)) (unless arg (funcall message-cite-function) 
(unless ... ...)) (case message-cite-reply-position (above ... ... ... ...) 
(below ...)) (unless modified (setq message-checksum ...)))
  (let nil (when (and message-reply-buffer message-cite-function) (when ... 
...) (if ... ...) (push-mark ...) (unless arg ... ...) (case message-cite-
reply-position ... ...) (unless modified ...)))
  eval((let nil (when (and message-reply-buffer message-cite-function) (when 
... ...) (if ... ...) (push-mark ...) (unless arg ... ...) (case message-
cite-reply-position ... ...) (unless modified ...))))
  message-yank-original()
  gnus-inews-yank-articles((56567))
  gnus-summary-reply((56567) t (56567))
  gnus-summary-very-wide-reply-with-original(nil)
  call-interactively(gnus-summary-very-wide-reply-with-original nil nil)

Re: case's definition is void when replying to mail

[Lars Magne Ingebrigtsen]

Raphael Kubo da Costa <kubito@gmail.com> writes:

> When using "S V" on a random IMAP message:
>
> Debugger entered--Lisp error: (void-function case)

Are you compiling Gnus with the same version of Emacs that you're
running it with?

It looks to me like the Emacs that compiled Gnus didn't know that `case'
was a macro.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: case's definition is void when replying to mail

[Raphael Kubo da Costa]

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

> Raphael Kubo da Costa <kubito <at> gmail.com> writes:
> 
> > When using "S V" on a random IMAP message:
> >
> > Debugger entered--Lisp error: (void-function case)
> 
> Are you compiling Gnus with the same version of Emacs that you're
> running it with?
> 
> It looks to me like the Emacs that compiled Gnus didn't know that `case'
> was a macro.

Yes, in this case it's this Emacs:
http://www.archlinux.org/packages/community/x86_64/emacs-nox, built this way:
http://projects.archlinux.org/svntogit/community.git/tree/emacs-nox/repos/community-x86_64/PKGBUILD

And whenever I fetch/merge from git, I always do `make clean && ./configure &&
make'.