Doesn't handle multiple emacs processes being open

[asmeurer]

I regularly use multiple emacs processes (in different terminal tabs), and this causes emacs to constantly complain about the keyfreq file being locked.

[dacap]

This is strange because keyfreq is already prepared to work with multiple emacs processes running in parallel (sometime I commonly do on Windows). Maybe the problem is when a lot of multiples instances are running. How many instances are you using?

I suspect that the problem is in keyfreq-table-save after keyfreq-file-is-unlocked is called. Could you debug it?

[asmeurer]

I haven't been using this since I reported this issue for this reason, but as I recall two instances were enough.

[asmeurer]

Hmm, so I'm not sure how to debug this. I have started seeing the issue again. It comes up during autosave. Is there a way to conditionally enable debug on quit only for that one message?

[swarnendubiswas]

I have faced the same issue and I had three instances open. In fact, enabling keyfreq package did not allow my emacs to quit with C-x C-c. Searching, I found that deleting the keyfreq stat files helped. There were two: one in home with suffix .lock, and the other was the stats file in ~/.emacs.d. Unfortunately I am not certain which file was the culprit. The parens were balanced in the stats file, whereas the .lock file was empty.

I have the same problem on another system. I will check that as well later today.

[dacap]

This problem is still happening in some situations.

[dacap]

Sorry again, but it looks that this issue isn't 100% fixed. I've asked to the Emacs mailing list about this:
http://lists.gnu.org/archive/html/help-gnu-emacs/2015-09/msg00359.html

[zimoun]

I have play around and it seems a classical multi-threading issue ?

Let two instances of Emacs, E1 and E2. And basically, when saving the scenario is:

• E1 checks if the file is locked.
• it is not, i.e. the call to keyfreq-file-is-unlocked l.472 from E1 returns t
• therefore E1 tries to lock it and call keyfreq-file-claim-lock.
• between these two calls by E1, the instance E2 wrote and locked, and therefore E1 complains it cannot write.

What I do not understand is: in the function keyfreq-file-claim-lock, the call write-region is inside ignore-error, so why Emacs is complaining ?

I have not found the option to tell to Emacs: try to write (emacs-pid) and if you cannot, no complain and pass.

Is it possible ?