Re: Two issues with the new Flymake

[Stefan Monnier]

[ Please report such problems via M-x report-emacs-bug, so they get
  a bug-number.  ]

> The first problem is something that happens when I add new code at the end of
> the source: if I write something and then quickly enough partially delete it
> with `delete-backward-char' and replace it with something else, at times
> Flymake reports an error message about unexpected report from the backend; I
> think it's due to the external checker reporting a problem in a region of the
> buffer that isn't there anymore at the time Flymake receives/processes that
> annotation. This is at worst a little annoyance, because at that time Emacs
> "hangs" for a noticeable interval of time, enough to disrupt my (re)writing.

What is the exact message you get?

> The second one is a strange interaction with the `python-shell-send-region' (a
> feature that I use very seldom): when I execute that, I get the following
> error message

>    Error in post-command-hook (#[0
>    "\304\305\303\242\306#\210r\302q\210\307\310\311\301\"\300\")\207" [nil
>    (post-command on-display) #<killed buffer> (#0) remove-hook
>    post-command-hook nil flymake-start remove post-command] 4]): (error
>    "Selecting deleted buffer")

This #[...] thingy above is the

        ((start-post-command
          ()
          (remove-hook 'post-command-hook #'start-post-command
                       nil)
          (with-current-buffer buffer
            (flymake-start (remove 'post-command deferred) force)))

function from flymake-start and the error comes from with-current-buffer
because `buffer` is not live any more when that code is run.

Not sure why that would happen, but here's my guess:

python-shell-send-region calls python-shell-buffer-substring
which does:

    (with-temp-buffer
      (python-mode)
      [...]

so it generates a temp buffer, puts it in python-mode, which in turns
runs python-mode-hook and hence enables flymake-mode which then
registers itself on post-command-hook.  But of course, by the time
post-command-hook is run, that temp buffer has been killed.

So I think the patch below should fix this.  I pushed it to `emacs-26`.


        Stefan


diff --git a/lisp/progmodes/flymake.el b/lisp/progmodes/flymake.el
index 1048bc5065..649e954893 100644
--- a/lisp/progmodes/flymake.el
+++ b/lisp/progmodes/flymake.el
@@ -742,8 +742,9 @@ flymake-start
           ()
           (remove-hook 'post-command-hook #'start-post-command
                        nil)
-          (with-current-buffer buffer
-            (flymake-start (remove 'post-command deferred) force)))
+          (when (buffer-live-p buffer)
+            (with-current-buffer buffer
+              (flymake-start (remove 'post-command deferred) force))))
          (start-on-display
           ()
           (remove-hook 'window-configuration-change-hook #'start-on-display