* martin rudalics [2010-06-20 12:42+0200] writes:

>> I'm wondering a bit though: kill-buffer protects the current buffer
>> while calling kill-buffer-hook.  Wouldn't it be prudent to do that for
>> the entire function?
>
> You mean for the case where the buffer we want to kill is not the
> current buffer?

Both cases.  Even in the case where the to-be-killed buffer is not
current, kill-buffer may decide not to actually kill it and instead
return nil; this case should also not switch the buffer.

>> Actually, I think there is a small bug there: if kill-buffer-hook is a
>> list of functions, the first function could potentially switch buffer
>> and the second function would be called in the wrong buffer.
>
> I suppose you're right.  Could you propose a patch for this and the
> above issue?

Below are two patches.  The first essentially adds a save-excursion
around the whole function.

The second patch adds a function run_hook_in_buffer. It iterates over
the functions in the hook and for each function explicitly sets the
buffer before calling it.  I had to introduce a new macro
DO_HOOK_FUNCTIONS which is a big hammer for this, but I couldn't find a
better way.

Helmut