Re: Fixing numerous `message' bugs..

[Glenn Morris]

Stefan Monnier wrote:

> The problem is that all those bugs are here because the corresponding
> authors just didn't think, so they are unlikely to use this new
> arg either.

Is there any way to improve byte-compile-format-warn in this regard?

It's easy to make it warn if the first arg to format is not a fixed
string or nil, but that leaves false positives for things like:

  (setq fmt "%s")
  (message fmt some-variable)

Can the byte-compiler figure out what is a known constant at compile
time? One would only want it to warn where the first argument is not
fully known at compile-time.

Because everyone pays attention to compiler warnings, right? ;)