Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[Alan Mackenzie]

Hi, Andreas!

On Tue, Jan 08, 2008 at 06:13:40PM +0100, Andreas Schwab wrote:
> c-{beginning,end}-of-defun try to modify the buffer and thus trigger
> ask-user-about-supersession-threat.  To reproduce load a C file and
> touch the underlying file.  Then type C-M-e repeatedly.  After some
> tries you will be asked whether you want to edit the buffer.

Ouch!

Yes, I can reproduce this.  At a guess, it's happening because of CC
Mode's extensive use of text properties.  I don't (yet) see why it only
happens after a several C-M-e's rather than the first.  Applying text
properties counts as modifying the buffer (I think this is a Bad Thing),
and the primitive function which applies the text properties seems to be
checking for the file having been modified.  Or something like that.

As I say, that's just a guess.  I'll take a closer look at it sometime.
I suspect it is one of these irritating little bugs whose solution
demands extensive redesign - here, perhaps, by using overlays/extents
instead of text properties.

I've added emacs-devel to the CC:

Thanks for the report!

> Andreas.

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[Lennart Borgman (gmail)]

Alan Mackenzie wrote:
> Hi, Andreas!
> 
> On Tue, Jan 08, 2008 at 06:13:40PM +0100, Andreas Schwab wrote:
>> c-{beginning,end}-of-defun try to modify the buffer and thus trigger
>> ask-user-about-supersession-threat.  To reproduce load a C file and
>> touch the underlying file.  Then type C-M-e repeatedly.  After some
>> tries you will be asked whether you want to edit the buffer.
> 
> Ouch!
> 
> Yes, I can reproduce this.  At a guess, it's happening because of CC
> Mode's extensive use of text properties.  I don't (yet) see why it only
> happens after a several C-M-e's rather than the first.  Applying text
> properties counts as modifying the buffer (I think this is a Bad Thing),
> and the primitive function which applies the text properties seems to be
> checking for the file having been modified.  Or something like that.

Are you talking about the text properties that font lock applies? I 
believe those does not mark the buffer as modified.

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[martin rudalics]

> As I say, that's just a guess.  I'll take a closer look at it sometime.
> I suspect it is one of these irritating little bugs whose solution
> demands extensive redesign - here, perhaps, by using overlays/extents
> instead of text properties.

Do you bind `buffer-file-name' to nil when applying text properties?

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[Lennart Borgman (gmail)]

martin rudalics wrote:
>> As I say, that's just a guess.  I'll take a closer look at it sometime.
>> I suspect it is one of these irritating little bugs whose solution
>> demands extensive redesign - here, perhaps, by using overlays/extents
>> instead of text properties.
> 
> Do you bind `buffer-file-name' to nil when applying text properties?


There is a macro in font-lock.el called save-buffer-state that font lock 
uses when it changes the text properties. Unfortunately that macro is 
not available for anyone else to use since it is in an 
eval-when-compile. Maybe that macro should be made available outside of 
font-lock.el -- for changing text properties?

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[martin rudalics]

> There is a macro in font-lock.el called save-buffer-state that font lock 
> uses when it changes the text properties. Unfortunately that macro is 
> not available for anyone else to use since it is in an 
> eval-when-compile. Maybe that macro should be made available outside of 
> font-lock.el -- for changing text properties?

Apparently people don't want it, see:

http://lists.gnu.org/archive/html/bug-gnu-emacs/2007-08/msg00068.html

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[Lennart Borgman (gmail)]

martin rudalics wrote:
>> There is a macro in font-lock.el called save-buffer-state that font 
>> lock uses when it changes the text properties. Unfortunately that 
>> macro is not available for anyone else to use since it is in an 
>> eval-when-compile. Maybe that macro should be made available outside 
>> of font-lock.el -- for changing text properties?
> 
> Apparently people don't want it, see:
> 
> http://lists.gnu.org/archive/html/bug-gnu-emacs/2007-08/msg00068.html

You mean because no one has responded to your request?

I believe I too have made the same request before, but with no response. 
If used properly I see no disadvantages with the save-buffer-state macro 
-- but otherwise it is of course dangerous.

However I do not know the actual reason that it is not made available. 
Could someone please explain?

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[martin rudalics]

 > You mean because no one has responded to your request?

Because I forgot to add this to `ada-set-syntax-table-properties'
then, despite asking for it ...

`antlr-mode' and `cperl-mode' knew better.


-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[Alan Mackenzie]

Hi, Martin!

On Tue, Jan 08, 2008 at 09:17:43PM +0100, martin rudalics wrote:
> > As I say, that's just a guess.  I'll take a closer look at it sometime.
> > I suspect it is one of these irritating little bugs whose solution
> > demands extensive redesign - here, perhaps, by using overlays/extents
> > instead of text properties.

> Do you bind `buffer-file-name' to nil when applying text properties?

No!  :-)

Thank you muchly for the tip!

I use a macro called c-save-buffer-state, which was modelled on
font-lock's original save-buffer-state some years ago.  At some stage,
presumably s-b-s got modified but c-s-b-s didn't.  Time to fix it!

I can't understand why save-buffer-state has never become a standard
Emacs macro.  If you're doing _anything_ with text properties, you can't
do without it.  (OK, Eli, there's one or two exceptions, I know! ;-)

-- 
Alan Mackenzie (Nuremberg, Germany).

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[Alan Mackenzie]

Hi, Andreas!

On Tue, Jan 08, 2008 at 06:13:40PM +0100, Andreas Schwab wrote:
> c-{beginning,end}-of-defun try to modify the buffer and thus trigger
> ask-user-about-supersession-threat.  To reproduce load a C file and
> touch the underlying file.  Then type C-M-e repeatedly.  After some
> tries you will be asked whether you want to edit the buffer.

Would you try the following patch, please.  Because it modifies a macro,
you'll need to byte-compile all the cc-*.el files before it will take
effect.


*** cc-defs.el~	2007-08-17 22:42:52.000000000 +0000
--- cc-defs.el	2008-01-09 08:57:19.414464408 +0000
***************
*** 425,430 ****
--- 425,431 ----
  	  (inhibit-read-only t) (inhibit-point-motion-hooks t)
  	  before-change-functions after-change-functions
  	  deactivate-mark
+ 	  buffer-file-name ; Prevent primitives checking for file modification
  	  ,@varlist)
       (unwind-protect
  	 (progn ,@body)


Let me know whether it works or not.

Thanks again for the report.

> Andreas.
 
-- 
Alan Mackenzie (Nuremberg, Germany).

Re: CC mode 5.31.4: c-{beginning, end}-of-defun try to modify buffer

[martin rudalics]

Please bind buffer-file-truename too.

> *** cc-defs.el~	2007-08-17 22:42:52.000000000 +0000
> --- cc-defs.el	2008-01-09 08:57:19.414464408 +0000
> ***************
> *** 425,430 ****
> --- 425,431 ----
>   	  (inhibit-read-only t) (inhibit-point-motion-hooks t)
>   	  before-change-functions after-change-functions
>   	  deactivate-mark
> + 	  buffer-file-name buffer-file-truename ; Prevent primitives checking for file modification
>   	  ,@varlist)
>        (unwind-protect
>   	 (progn ,@body)


-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace