Re: ediff failure if diff doesn't exist

[Kevin Rodgers]

Jesper Harder wrote:
> On ms-windows when you (require 'ediff) on a computer where the diff
> executable doesn't exist you get the backtrace below.
>
> Of course, ediff probably isn't that useful if you don't have diff,
> but still it would be nice to fail in a more gracefull way.
>
> To reproduce, set ediff-diff-program to something that doesn't exist
> and require ediff.
>
> Debugger entered--Lisp error: (file-error "Searching for program" "no
> such file or directory" "diff")
>  call-process("diff" nil nil nil "--binary" "NUL" "NUL")
>  apply(call-process ("diff" nil nil nil "--binary" "NUL" "NUL"))
>  ediff-diff-mandatory-option("diff")
>  ediff-reset-diff-options(ediff-diff-options "")
>  custom-initialize-reset(ediff-diff-options "")
>  custom-declare-variable(ediff-diff-options ""
> ("r:/Dev/Common/Util/Emacs/lisp/ediff-diff.elc" . -3371) :set
> ediff-reset-diff-options :type string :group ediff-diff)
>  require(ediff-diff)

I came across a similar problem in one of my own packages that uses
call-process to determine what options an external program supports.

It's easy to fix by checking first that executable-find returns non-nil,
or by wrapping call-process in condition-case.

But that wasn't necessary before, so I suspect that call-process has
changed to now signal an error.  If that's intentional, its doc string
should be updated, because currently it says only this:

| If BUFFER is 0, `call-process' returns immediately with value nil.
| Otherwise it waits for PROGRAM to terminate
| and returns a numeric exit status or a signal description string.
| If you quit, the process is killed with SIGINT, or SIGKILL if you quit 
again.

--
Kevin