Confusion about first-change-hook

Confusion about first-change-hook

[David Engster]

I wanted to do something very simple: Activate whitespace-mode only when
I start editing a buffer. So I looked into the manual and found
`first-change-hook', which seemed exactly what I wanted.

However, I soon saw that whitespace-mode got immediately activated in my
C++ buffers. The reason is that changing a text property apparently
already counts as "changing the buffer". Is this really the desired
behavior? If so, what else is there to achieve the above? Should I file
a wishlist-bug for something like 'buffer-modified-hook'?

-David

Re: Confusion about first-change-hook

[Stefan Monnier]

> However, I soon saw that whitespace-mode got immediately activated in my
> C++ buffers. The reason is that changing a text property apparently
> already counts as "changing the buffer". Is this really the desired
> behavior? If so, what else is there to achieve the above? Should I file
> a wishlist-bug for something like 'buffer-modified-hook'?

I think the bug is in the major mode that sets the text-property without
proper protection (e.g. with-silent-modifications).


        Stefan

Re: Confusion about first-change-hook

[David Engster]

Stefan Monnier writes:
>> However, I soon saw that whitespace-mode got immediately activated in my
>> C++ buffers. The reason is that changing a text property apparently
>> already counts as "changing the buffer". Is this really the desired
>> behavior? If so, what else is there to achieve the above? Should I file
>> a wishlist-bug for something like 'buffer-modified-hook'?
>
> I think the bug is in the major mode that sets the text-property without
> proper protection (e.g. with-silent-modifications).

Oh. I didn't know that putting text properties on an unmodified buffer
should be protected like that. I guess I will have to look at some of my
own packages...

Anyway, thanks for the pointer. I'll see to it that I file a bug for
cc-mode.

-David

Re: Confusion about first-change-hook

[Jambunathan K]

David Engster <deng@randomsample.de> writes:

> I wanted to do something very simple: Activate whitespace-mode only when
> I start editing a buffer. So I looked into the manual and found
> `first-change-hook', which seemed exactly what I wanted.

This seems to work:

    (add-hook 'c++-mode-hook
              (lambda nil
                (add-hook 'before-change-functions 'activate-whitespace-mode
                          nil 'local)))

    (defun activate-whitespace-mode (beg end) 
      (whitespace-mode 1)
      (remove-hook 'before-change-functions
                   'activate-whitespace-mode 'local))

>
> However, I soon saw that whitespace-mode got immediately activated in my
> C++ buffers. The reason is that changing a text property apparently
> already counts as "changing the buffer". Is this really the desired
> behavior? If so, what else is there to achieve the above? Should I file
> a wishlist-bug for something like 'buffer-modified-hook'?
>
> -David
>
>

-- 

Re: Confusion about first-change-hook

[David Engster]

Jambunathan K. writes:
> David Engster <deng@randomsample.de> writes:
>
>> I wanted to do something very simple: Activate whitespace-mode only when
>> I start editing a buffer. So I looked into the manual and found
>> `first-change-hook', which seemed exactly what I wanted.
>
> This seems to work:
>
>     (add-hook 'c++-mode-hook
>               (lambda nil
>                 (add-hook 'before-change-functions 'activate-whitespace-mode
>                           nil 'local)))
>
>     (defun activate-whitespace-mode (beg end) 
>       (whitespace-mode 1)
>       (remove-hook 'before-change-functions
>                    'activate-whitespace-mode 'local))

Thank you. This works nicely.

-David

Re: Confusion about first-change-hook

[Stefan Monnier]

> Oh. I didn't know that putting text properties on an unmodified buffer
> should be protected like that. I guess I will have to look at some of my
> own packages...

Many packages do part of what with-silent-modifications does, such as
save buffer-modified-p and restore it afterwards, but that is only
sufficient in some cases (as seen in your particular case).
`with-silent-modifications' takes care of several more potential
problems.


        Stefan