Re: ert.el --- Emacs Lisp Regression Testing

[Christian Ohler <ohler+emacs@fastmail.net>]

Lennart Borgman (gmail), 2007-12-31:

> I will try to test it ;-)

Please do so -- perhaps by writing a few test cases for Emacs ;-)


> One thing that might be handy is something that runs selected entries in 
> a new instance of Emacs, starting from either "emacs" or "emacs -Q". I 
> think that is one of the most common things I do when I am testing.

ERT currently allows you to run your tests in a fresh emacs by running

     emacs -Q --batch --load my-tests.el --eval '(ert-run-tests-batch t)'

in a terminal (or possibly under M-x compile, but I haven't tried that).

However, that would be an edit-compile-run development cycle.  I am much 
more productive programming and testing interactively in one single 
Emacs instance.

I see two main reasons for running tests in a separate Emacs: For 
testing without customizations (emacs -Q) or other Emacs versions, and 
for long-running test suites that you don't want to block your main 
Emacs instance with.  So far, the above method was sufficient for me in 
those situations (because tests that have already succeeded in my Emacs 
rarely fail in a separate one, so I rarely missed ERT's interactive 
debugging facilities).

Maybe IELM could be used to implement a convenient interface for running 
an inferior Emacs that runs the tests (which could, in turn, involve 
running another nested Emacs in M-x term).  This even sounds like an 
interesting project to implement.  However, to get started building test 
suites for Emacs or Elisp packages, the current options are probably 
sufficient.

Christian.