ERC tracking broken

ERC tracking broken

[Antoine Levitt]

ERC tracking seems to be broken. Reverting to the binary of emacs I
compiled yesterday fixes it, so I strongly suspect it's a result of

commit 5e4cb836580a8ef2f9ea9b4b8317005941920027
Author: Stefan Monnier <monnier@iro.umontreal.ca>
Date:   Mon Mar 28 16:26:35 2011 -0400

    Don't reset post-command-hook to nil upon error.

which interferes with a patch by Julien Danjou about using
post-command-hook to notice window changes in erc-track. I don't
remember the specifics though.

Also, it seems it has broken ido completion, but I'm not sure exactly
how and it might be related to custom settings I have. The ERC thing
seems easier to debug.

Re: ERC tracking broken

[Stefan Monnier]

> ERC tracking seems to be broken. Reverting to the binary of emacs I
> compiled yesterday fixes it, so I strongly suspect it's a result of

> commit 5e4cb836580a8ef2f9ea9b4b8317005941920027
> Author: Stefan Monnier <monnier@iro.umontreal.ca>
> Date:   Mon Mar 28 16:26:35 2011 -0400
>     Don't reset post-command-hook to nil upon error.

> which interferes with a patch by Julien Danjou about using
> post-command-hook to notice window changes in erc-track.  I don't
> remember the specifics though.

My change should not affect the behavior of post-command-hook expect
when one of the hook functions signals an error (something that should
not happen, normally), so if it's due to my change, it's a bug in
my change.

Do you have a recipe to reproduce the problem, starting from "emacs -Q"
(including for someone like me who basically never used IRC)?


        Stefan

Re: ERC tracking broken

[Antoine Levitt]

29/03/11 23:20, Stefan Monnier
>> ERC tracking seems to be broken. Reverting to the binary of emacs I
>> compiled yesterday fixes it, so I strongly suspect it's a result of
>
>> commit 5e4cb836580a8ef2f9ea9b4b8317005941920027
>> Author: Stefan Monnier <monnier@iro.umontreal.ca>
>> Date:   Mon Mar 28 16:26:35 2011 -0400
>>     Don't reset post-command-hook to nil upon error.
>
>> which interferes with a patch by Julien Danjou about using
>> post-command-hook to notice window changes in erc-track.  I don't
>> remember the specifics though.
>
> My change should not affect the behavior of post-command-hook expect
> when one of the hook functions signals an error (something that should
> not happen, normally), so if it's due to my change, it's a bug in
> my change.
>
> Do you have a recipe to reproduce the problem, starting from "emacs -Q"
> (including for someone like me who basically never used IRC)?
>
>
>         Stefan

Alright, I finally traced it back to something not ERC-related. The
easiest way I could reproduce it is with some code to highlight
parentheses. So here's a simple recipe:

- Get http://pastebin.com/UafFZ9qy and put it in your load path (I got
  this code from emacswiki some time ago, it worked fine until today -
  which doesn't mean it's correct, maybe your changes simply exposed a
  bug in it)

- Put

(require 'highlight-parentheses)
(global-highlight-parentheses-mode t)
(ido-mode 1)
(icomplete-mode 1)

in a blank .emacs

- M-x a, observe there's a nice completion list
- C-g to get out of the completion
- M-x a, observe the completion list has disappeared (for the rest
  of the session)

This started occuring at some point today and the code uses
post-command-hook, so it's probably related.

Hope that helps

Antoine

Re: ERC tracking broken

[Stefan Monnier]

> Alright, I finally traced it back to something not ERC-related. The
> easiest way I could reproduce it is with some code to highlight
> parentheses. So here's a simple recipe:

> - Get http://pastebin.com/UafFZ9qy and put it in your load path (I got
>   this code from emacswiki some time ago, it worked fine until today -
>   which doesn't mean it's correct, maybe your changes simply exposed a
>   bug in it)

> - Put

> (require 'highlight-parentheses)
> (global-highlight-parentheses-mode t)
> (ido-mode 1)
> (icomplete-mode 1)

> in a blank .emacs

> - M-x a, observe there's a nice completion list
> - C-g to get out of the completion
> - M-x a, observe the completion list has disappeared (for the rest
>   of the session)

I did:

% trunk/src/emacs -Q -l /.../highlight-parentheses.el \
                     -f global-highlight-parentheses-mode \
                     -f ido-mode \
                     -f icomplete-mode
M-x a, observe the completion list
C-g to get out of M-x
M-x a

and I observed the completion list again.
I.e. I can't reproduce it here :-(


        Stefan


PS: Why use ido and icomplete?  Doesn't ido already do what icomplete does?

Re: ERC tracking broken

[Antoine Levitt]

30/03/11 02:56, Stefan Monnier
>> Alright, I finally traced it back to something not ERC-related. The
>> easiest way I could reproduce it is with some code to highlight
>> parentheses. So here's a simple recipe:
>
>> - Get http://pastebin.com/UafFZ9qy and put it in your load path (I got
>>   this code from emacswiki some time ago, it worked fine until today -
>>   which doesn't mean it's correct, maybe your changes simply exposed a
>>   bug in it)
>
>> - Put
>
>> (require 'highlight-parentheses)
>> (global-highlight-parentheses-mode t)
>> (ido-mode 1)
>> (icomplete-mode 1)
>
>> in a blank .emacs
>
>> - M-x a, observe there's a nice completion list
>> - C-g to get out of the completion
>> - M-x a, observe the completion list has disappeared (for the rest
>>   of the session)
>
> I did:
>
> % trunk/src/emacs -Q -l /.../highlight-parentheses.el \
>                      -f global-highlight-parentheses-mode \
>                      -f ido-mode \
>                      -f icomplete-mode

This is fine

> M-x a, observe the completion list
> C-g to get out of M-x
> M-x a
>
> and I observed the completion list again.
> I.e. I can't reproduce it here :-(

That's weird. I just checked using a make bootstrap, the problem is
still there, at latest bzr.

>
>
>         Stefan
>
>
> PS: Why use ido and icomplete?  Doesn't ido already do what icomplete does?

No, icomplete adds "inline" completion, which icomplete doesn't
do. Maybe I wasn't too clear earlier: when I was speaking of the
completion list, I meant the completion list from icomplete that appears
in real time in the minibuffer when typing things (which probably uses
post-command-hook), not the one you get by pressing TAB (which probably
doesn't). Are we talking about the same thing?

PS: actually, this is slighty more subtle than I thought. Try the
following, still with the same initialisation

M-x C-g
M-x t (inline completion on) o (inline completion off, which is a bug)
TAB (inline completion on)

Antoine

Re: ERC tracking broken

[Stefan Monnier]

> - Get http://pastebin.com/UafFZ9qy and put it in your load path (I got

For next time(s), please give a URL to the actual code rather than to an
HTML rendition of it, so I don't have to go through a browser.

> This started occuring at some point today and the code uses
> post-command-hook, so it's probably related.

This should now be fixed,


        Stefan