Re: [PATCH] make test: Make failure results more verbose

[Max Nikulin]

On 02/01/2022 20:12, Ihor Radchenko wrote:
> In newer Emacs, ERT is capable of providing more info about FAILED
> tests. Maybe we can enable this option by default in the Org test suite?
Thinking more, I have realized that something is wrong.

Behavior of tests in Org should be controlled by EMACS_TEST_VERBOSE 
*environment* variable. Org makefiles may have it on by default, but it 
should not override explicitly chosen value. Fortunately Emacs 
repository contains no makefiles from Org, so it will not cause conflict 
with Emacs builds.

The problem is that EMACS_TEST_VERBOSE interface is broken. In the 
following case I expect that test summary should not be verbose:

ert-sample.el
---- >8 ----
(require 'ert)
(ert-deftest ert-sample ()
  (should (equal 5 (* 2 2))))
(ert-run-tests-batch-and-exit)
---- 8< ----

(I am tried it with Emacs-27, so I put a copy of ert.el from git HEAD to 
~/ert)

EMACS_TEST_VERBOSE= emacs --batch -Q -L ~/ert -l ert-sample.el

1 unexpected results:
   FAILED  ert-sample  ((should (equal 5 (* 2 2))) :form (equal 5 4) 
:value nil :explanation (different-atoms (5 "#x5" "?") (4 "#x4" "?")))

That is why I am going to file a bug against ert.

In a minimal variant empty string (`getenv' return value) should not be 
considered as t.

For better user experience, I expect that the following case-insensitive 
strings should be considered as false: "0", "false", "no", "n", "off", 
"f" (borrowed from .ini), "none", "nil". Other values should be 
considered as true or anything besides "1", "true", "yes", "y", "on", 
"t" should be considered as invalid value.

I am attaching a tentative patch for Org that should make 
EMACS_TEST_VERBOSE setting more transparent.

0001-make-test-Make-failure-summary-more-verbose.patch
Description: Text Data