font-lock and shell scripts: Raises redisplay errors in highlight pattern:

font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[T. V. Raman]

Hi,

This has been around for some months now and I've been too lazy
to track down its exact source until now.

Symptom:

When editting shell scripts with font-lock turned on, emacs
raises errors during redisplay -- specifically in functions 

font-lock-apply-syntactic-highlight
font-lock-apply-highlight

Once these errors hit, highlighting doesn't appear to happen
correctly.

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[Romain Francoise]

"T. V. Raman" <raman@users.sf.net> writes:

> When editting shell scripts with font-lock turned on, emacs
> raises errors during redisplay -- specifically in functions

> font-lock-apply-syntactic-highlight
> font-lock-apply-highlight

> Once these errors hit, highlighting doesn't appear to happen
> correctly.

Something in your configuration is adding invalid entries to the
`font-lock-keywords' variable and it confuses the font-lock engine.
I'm pretty sure that there is no bug in Emacs here.  Could you please
try to track down which package or init file changes the font lock
keywords for this mode in your configuration?

Thanks,

-- 
Romain Francoise <romain@orebokech.com> | The sea! the sea! the open
it's a miracle -- http://orebokech.com/ | sea! The blue, the fresh, the
                                        | ever free! --Bryan W. Procter

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[Kim F. Storm]

"T. V. Raman" <raman@users.sf.net> writes:

> Hi,
>
> This has been around for some months now and I've been too lazy
> to track down its exact source until now.
>
> Symptom:
>
> When editting shell scripts with font-lock turned on, emacs
> raises errors during redisplay -- specifically in functions 
>
> font-lock-apply-syntactic-highlight
> font-lock-apply-highlight
>
> Once these errors hit, highlighting doesn't appear to happen
> correctly.

What kind of errors are raised "during redisplay".

Does turning on Options > Debug on Error provide more information?

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[T. V. Raman]

I'll try to reproduce a backtrace --- wil need to first disable
the advice I threw around it to get rid of the annoyance.

Note that I've not changed or modified font-lock-keywords for
shell-script mode as far as I can tell.

One possibility is that I'm noticing these errors when others
dont because I advice 'error in emacspeak to get auditory alerts
for errors.
Since 'error does not return, this has to be a before advice.
A side-consequence of this is that errors that are later by  an
error-handler further up the call stack end up still getting
signaled auditorally. 




>>>>> "Kim" == Kim F Storm <storm@cua.dk> writes:
    Kim> "T. V. Raman" <raman@users.sf.net> writes:
    >> Hi,
    >> 
    >> This has been around for some months now and I've been too
    >> lazy to track down its exact source until now.
    >> 
    >> Symptom:
    >> 
    >> When editting shell scripts with font-lock turned on,
    >> emacs raises errors during redisplay -- specifically in
    >> functions
    >> 
    >> font-lock-apply-syntactic-highlight
    >> font-lock-apply-highlight
    >> 
    >> Once these errors hit, highlighting doesn't appear to
    >> happen correctly.
    Kim> 
    Kim> What kind of errors are raised "during redisplay".
    Kim> 
    Kim> Does turning on Options > Debug on Error provide more
    Kim> information?
    Kim> 
    Kim> -- Kim F. Storm <storm@cua.dk> http://www.cua.dk

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[Stefan Monnier]

> One possibility is that I'm noticing these errors when others
> dont because I advice 'error in emacspeak to get auditory alerts
> for errors.
> Since 'error does not return, this has to be a before advice.
> A side-consequence of this is that errors that are later by  an
> error-handler further up the call stack end up still getting
> signaled auditorally. 

Please post a feature request with M-x report-emacs-bug:
it should be possible to customize the top-level handling of errors
(currently hardcoded in cmd_error_internal).


        Stefan

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[T. V. Raman]

Here are some more details on this problem.

The error is definitely being raised -- but also caught by the
condition-case  handler in functions
font-lock-default-fontify-buffer and
font-lock-default-fontify-region.

The condition-case in those functions means that
toggle-debug-on-error does not give a backtrace alas.

But I am able to reliably trigger it as follows:

Open an empty file and put it in sh-mode.

Type:

#!/bin/bash
export a=1

You see the error immediately after you hit the ' ' after the
keyword "export".

The error being raised  (from searching the font-lock.el sources,
it's one of the font-lock-apply-* functions)
Error during redisplay: (error No match 6 in highlight (6
font-lock-builtin-face))



So two questions:

1) Might be marginally benefitial even for the mainstream emacs user
to not have these errors raised and handled --
--- what variables in my sh-mode/font-lock environment should I
look at to chase down and fix the cause of the error in the first
place? (Note that I'm running with font-lock turned on at the
highest level).

2) And I'll possibly file this as a separate bug report once
   Stephane elaborates on what he said toward the end of his
   note:


How can I tell programmatically at run time if an error that is
raised has a handler somewhere above in the call stack waiting to
handle it?

If I had (2), then my advice  on error could be smart in the
following way:

When (error ...) is called:

Before-Advice:

A)      Check if there is a handler waiting to catch the error.
B)      If yes, do nothing;
C) Else give the user auditory feedback.

I could achieve an equivalent effect if I  knew how to check the
following in elisp:

Check if an error has "bubbled" all the way up the call stack.



Note that I have to advice error in Emacspeak because too many
useful Emacs packages use (error ...) to signal useful
information ---as an example, VM (the mail reader I use)
uses error to signal end of folder.


would be nice if 

>>>>> "Stefan" == Stefan Monnier <monnier@iro.umontreal.ca> writes:
    >> One possibility is that I'm noticing these errors when
    >> others dont because I advice 'error in emacspeak to get
    >> auditory alerts for errors.  Since 'error does not return,
    >> this has to be a before advice.  A side-consequence of
    >> this is that errors that are later by an error-handler
    >> further up the call stack end up still getting signaled
    >> auditorally.
    Stefan> 
    Stefan> Please post a feature request with M-x
    Stefan> report-emacs-bug: it should be possible to customize
    Stefan> the top-level handling of errors (currently hardcoded
    Stefan> in cmd_error_internal).
    Stefan> 
    Stefan> 
    Stefan>         Stefan

 Thanks, 
 --Raman

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[Stefan Monnier]

> Error during redisplay: (error No match 6 in highlight (6
> font-lock-builtin-face))

Thanks.  Does the patch below fix your problem?


        Stefan


--- sh-script.el	04 Jul 2006 11:23:07 -0400	1.181
+++ sh-script.el	13 Jul 2006 11:28:47 -0400	
@@ -1554,7 +1554,7 @@
 			 (regexp-opt (sh-feature sh-builtins) t)
 			 "\\>")
 		(2 font-lock-keyword-face nil t)
-		(6 font-lock-builtin-face))
+		(4 font-lock-builtin-face))
 	       ,@(sh-feature sh-font-lock-keywords-var-2)))
 	 (,(concat keywords "\\)\\>")
 	  2 font-lock-keyword-face)

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[T. V. Raman]

Suspect your patch will fix it.
I did a cvs update before patching --- turned out to be a mistake
in hindsight because the build is currently broken like so:)
(this is after a make clean , configure ...

../src/emacs -batch --no-site-file --multibyte -f batch-byte-compile /home/tvraman/sourceforge/emacs/lisp/progmodes/cc-mode.el
Source file `/home/tvraman/sourceforge/emacs/lisp/emacs-lisp/edebug.el' newer than byte-compiled file
Source file `/home/tvraman/sourceforge/emacs/lisp/progmodes/cc-fonts.el' newer than byte-compiled file
cc-mode.el:1188:21:Error: Wrong type argument: integerp, nil
make: *** [/home/tvraman/sourceforge/emacs/lisp/progmodes/cc-mode.elc] Error 1
18:27:05 labrador lisp $ 

>>>>> "Stefan" == Stefan Monnier <monnier@iro.umontreal.ca> writes:
    >> Error during redisplay: (error No match 6 in highlight (6
    >> font-lock-builtin-face))
    Stefan> 
    Stefan> Thanks.  Does the patch below fix your problem?
    Stefan> 
    Stefan> 
    Stefan>         Stefan
    Stefan> 
    Stefan> 
    Stefan> --- sh-script.el 04 Jul 2006 11:23:07 -0400 1.181 +++
    Stefan> sh-script.el 13 Jul 2006 11:28:47 -0400 @@ -1554,7
    Stefan> +1554,7 @@ (regexp-opt (sh-feature sh-builtins) t)
    Stefan> "\\>") (2 font-lock-keyword-face nil t) - (6
    Stefan> font-lock-builtin-face)) + (4
    Stefan> font-lock-builtin-face)) ,@(sh-feature
    Stefan> sh-font-lock-keywords-var-2))) (,(concat keywords
    Stefan> "\\)\\>") 2 font-lock-keyword-face)

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[Richard Stallman]

    1) Might be marginally benefitial even for the mainstream emacs user
    to not have these errors raised and handled --
    --- what variables in my sh-mode/font-lock environment should I
    look at to chase down and fix the cause of the error in the first
    place? (Note that I'm running with font-lock turned on at the
    highest level).

If there is no good reason for this to signal an error, it would be
cleaner to eliminate it.  But first one has to track down what causes
the error.  One way to do that is using GDB with a breakpoint at
Fsignal.  Setting debug-on-signal might enable you to find the cause
using the Lisp debugger.

    When (error ...) is called:

    Before-Advice:

    A)      Check if there is a handler waiting to catch the error.
    B)      If yes, do nothing;
    C) Else give the user auditory feedback.

Unhandled errors do make a beep, which is auditory feedback.  So what
is the change that you propose?

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[T. V. Raman]

Unhandled errors do make a beep which is correct.

But that is not sufficient for the  user who is not looking at
the display --- I need to say "why" it beeped.

Also, very often I dont sit at my PC -- I sit somewhere outside
with a wireless keyboard and speakers, and sadly the PC speaker
beep only beeps at the PC -- not over the sound card.

The latter is a bit esoteric I admit -- but for something like
the vm mail reader signalling end of folder I really need more
feedback than a beep.


>>>>> "Richard" == Richard Stallman <rms@gnu.org> writes:
    Richard>     1) Might be marginally benefitial even for the
    Richard> mainstream emacs user to not have these errors
    Richard> raised and handled -- --- what variables in my
    Richard> sh-mode/font-lock environment should I look at to
    Richard> chase down and fix the cause of the error in the
    Richard> first place? (Note that I'm running with font-lock
    Richard> turned on at the highest level).
    Richard> 
    Richard> If there is no good reason for this to signal an
    Richard> error, it would be cleaner to eliminate it.  But
    Richard> first one has to track down what causes the error.
    Richard> One way to do that is using GDB with a breakpoint at
    Richard> Fsignal.  Setting debug-on-signal might enable you
    Richard> to find the cause using the Lisp debugger.
    Richard> 
    Richard>     When (error ...) is called:
    Richard> 
    Richard>     Before-Advice:
    Richard> 
    Richard>     A) Check if there is a handler waiting to catch
    Richard> the error.  B) If yes, do nothing; C) Else give the
    Richard> user auditory feedback.
    Richard> 
    Richard> Unhandled errors do make a beep, which is auditory
    Richard> feedback.  So what is the change that you propose?

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[David Kastrup]

"T. V. Raman" <raman@users.sf.net> writes:

> Unhandled errors do make a beep which is correct.
>
> But that is not sufficient for the  user who is not looking at
> the display --- I need to say "why" it beeped.
>
> Also, very often I dont sit at my PC -- I sit somewhere outside
> with a wireless keyboard and speakers, and sadly the PC speaker
> beep only beeps at the PC -- not over the sound card.

Could be a matter of mixer control.

> The latter is a bit esoteric I admit -- but for something like
> the vm mail reader signalling end of folder I really need more
> feedback than a beep.

visible-bell is a variable defined in `C source code'.
Its value is t

Documentation:
*Non-nil means try to flash the frame to represent a bell.

See also `ring-bell-function'.

You can customize this variable.

[back]


-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[Chong Yidong]

"T. V. Raman" <raman@users.sourceforge.net> writes:

> Also, very often I dont sit at my PC -- I sit somewhere outside
> with a wireless keyboard and speakers, and sadly the PC speaker
> beep only beeps at the PC -- not over the sound card.

This is tangential to the discussion, but:

  visible-bell is a variable defined in `C source code'.
  Its value is nil

  Documentation:
  *Non-nil means try to flash the frame to represent a bell.

Unless, of course, you sit outside without a *monitor* either.  Which
would be a bit like blind chess, I guess.

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[T. V. Raman]

blind computing actually. And yes, I sit without a monitor ---
why would I want to carry around a photon generator when I'm photon-independent?

>>>>> "Chong" == Chong Yidong <cyd@stupidchicken.com> writes:
    Chong> "T. V. Raman" <raman@users.sourceforge.net> writes:
    >> Also, very often I dont sit at my PC -- I sit somewhere
    >> outside with a wireless keyboard and speakers, and sadly
    >> the PC speaker beep only beeps at the PC -- not over the
    >> sound card.
    Chong> 
    Chong> This is tangential to the discussion, but:
    Chong> 
    Chong>   visible-bell is a variable defined in `C source
    Chong> code'.  Its value is nil
    Chong> 
    Chong>   Documentation: *Non-nil means try to flash the frame
    Chong> to represent a bell.
    Chong> 
    Chong> Unless, of course, you sit outside without a *monitor*
    Chong> either.  Which would be a bit like blind chess, I
    Chong> guess.

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[Richard Stallman]

    Unhandled errors do make a beep which is correct.

    But that is not sufficient for the  user who is not looking at
    the display --- I need to say "why" it beeped.

Do you have a change in Emacs to suggest to us?

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[T. V. Raman]

I would not make a general change to emacs for this.


For Emacspeak I'd like to have some mechanism that I can hook to
reliably --- perhaps an unhandled-error-hook-function that I can
latch some speech output to.

>>>>> "Richard" == Richard Stallman <rms@gnu.org> writes:
    Richard>     Unhandled errors do make a beep which is
    Richard> correct.  But that is not sufficient for the user
    Richard> who is not looking at the display --- I need to say
    Richard> "why" it beeped.
    Richard> 
    Richard> Do you have a change in Emacs to suggest to us?

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[Richard Stallman]

    I would not make a general change to emacs for this.

    For Emacspeak I'd like to have some mechanism that I can hook to
    reliably --- perhaps an unhandled-error-hook-function that I can
    latch some speech output to.

That IS a general change to Emacs.  (That is not a criticism.  We
often make general changes in Emacs.)

Would a hook in cmd_error in keyboard.c do what you have in mind?

Re: font-lock and shell scripts: Raises redisplay errors in highlight pattern:

[T. V. Raman]

I believe a hook as you describe hould be sufficient. 


The hook function should have access to the error message.



>>>>> "Richard" == Richard Stallman <rms@gnu.org> writes:
    Richard>     I would not make a general change to emacs for
    Richard> this.  For Emacspeak I'd like to have some mechanism
    Richard> that I can hook to reliably --- perhaps an
    Richard> unhandled-error-hook-function that I can latch some
    Richard> speech output to.
    Richard> 
    Richard> That IS a general change to Emacs.  (That is not a
    Richard> criticism.  We often make general changes in Emacs.)
    Richard> 
    Richard> Would a hook in cmd_error in keyboard.c do what you
    Richard> have in mind?

 Thanks, 
 --Raman

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs