Re: Avoid asking to user for lock (ask-user-about-lock)

[David Capello <davidcapello@gmail.com>]

> > The problem is here:
> > https://github.com/dacap/keyfreq/blob/master/keyfreq.el#L427
>
> I think rather than override ask-user-about-lock you want to use the
> MUSTBENEW argument:
>
>  (write-region (number-to-string (emacs-pid)) nil
>                keyfreq-file-lock nil 'nomessage nil nil 'excl)

I think there is one extra nil, so I've used

  (write-region (number-to-string (emacs-pid)) nil
                keyfreq-file-lock nil 'nomessage nil 'excl)

And it seems to work, but I still getting some messages like:

  .../.emacs.keyfreq.lock locked by david@pc (pid 9076): (s, q, p, ?)?

in the minibuf.

PS: Please CC me because I'm not subscribed to the list.

- David

On Thu, Sep 24, 2015 at 9:39 AM, David Capello <davidcapello@gmail.com>
wrote:

> Hi there,
>
> I've a package (keyfreq) that tries to add data to a file each X seconds.
> To do this it has to create a lock file, write the file, and release/delete
> the lock. If two instances of Emacs are running, both instances will try
> to write in the file at certain point. And here is where I'm having
> troubles
> (this was not a problem in Emacs 24.3, but now it is on 24.5).
>
> The problem is here:
> https://github.com/dacap/keyfreq/blob/master/keyfreq.el#L427
>
> I'm trying to redefine ask-user-about-lock temporarily so it doesn't
> ask to overwrite the .lock file if it already exists (if the lock already
> exist, we just "do nothing"/"wait next timeout", because the file is
> locked).
> Anyway it looks that sometimes, this redefinition doesn't work correctly.
>
> I would like to know if there is a way/common pattern to handle
> this kind of situations where two or more instances of Emacs are trying
> to add data to a file, and if they cannot get the lock, they just do
> nothing and can try it later.
>
> - David
>