On Tue, Sep 19, 2017 at 9:00 AM Stefan Monnier <monnier@iro.umontreal.ca>
wrote:

> > This was fallout from 9b980e269, aka bug #28412, where now we prompt for
> > buffers that have functions in the `write-contents-functions' hook.
>
> Ah, I see I was looking at the wrong version of the code, sorry.
> So the question becomes: why does " *Code-Fontify*" have a non-nil
> write-contents-functions?
>

It's because ggtags.el enables the major mode from the actual file in this
temp " *Code-Fontify*" buffer by calling (funcall mode) in
ggtags-fontify-code.

And in my case, the verilog-mode is setting the write-contents-hooks (old
name of write-contents-functions)[1].


Looking at ggtags.el, I also wonder why the buffer survives the call to
> ggtags-fontify-code.  I would have expected ggtags-fontify-code to look
> more like:
>
> (cl-defun ggtags-fontify-code (code &optional (mode major-mode))
>   (cl-check-type mode function)
>   (cl-typecase code        ;; FIXME: Using `stringp` is more efficient.
>     ((not string) code)
>     (string
>      (with-temp-buffer
>        (insert code)
>        ;; FIXME: `delay-mode-hooks' needs a comment
>        (delay-mode-hooks (funcall mode))
>        (font-lock-ensure)
>        (buffer-string)))))
>

I will start this discussion on the ggtags.el repo. Thanks.

[1]:
https://github.com/veripool/verilog-mode/blob/ecdaea3386bee9d38520914e458d4860da644773/verilog-mode.el#L3969

-- 

Kaushal Modi