On Thu, 29 Dec 2011 16:05:39 +0000, David Edmondson <dme@dme.org> wrote:
> What's the general feeling about this approach to producing tests for
> the emacs UI? (That is, code the test in a .el file and call the
> relevant function(s) from the test harness.)

I think it's a nice idea.  It seems much cleaner than writing ever more
complicated code in the command line.  There was another recent idea to
actually write the python tests in python, if I remember correctly.  It
all seems to make sense to me.  Bash is convenient, but there's no
reason we need to be monogamous, especially when there are ways, with
your technique, to call individual "embedded" tests from the command
line and still take advantage of the bells and whistles of full test
infrastructure.

> It makes it simpler to develop and maintain the test (because you can do
> more work with traditional emacs support for editing elisp), but might
> make interpreting failures more difficult (the test harness mostly just
> reports 'failed').

It is really nice to see diffs on failure, actually, to get a sense of
what exactly went wrong.  Is it possible to have some sort of standard
report to stdout that could provide more info?

jamie.