Re: emacs-30 6bca138d60e: Fix latest test for dabbrev-expand

[Stephen Berman]

On Mon, 02 Dec 2024 15:45:23 -0500 Andrea Corallo <acorallo@gnu.org> wrote:

> Stephen Berman via Mailing list for Emacs changes <emacs-diffs@gnu.org>
> writes:
>
>> branch: emacs-30
>> commit 6bca138d60e47d0fbab4faf8e83c5b9b319f260d
>> Author: Stephen Berman <stephen.berman@gmx.net>
>> Commit: Stephen Berman <stephen.berman@gmx.net>
>>
>>     Fix latest test for dabbrev-expand
>>
>>     * test/lisp/dabbrev-tests.el (dabbrev-expand-after-killing-buffer):
>>     Fix typo in a 'should' test, use part of return value of 'should-error'
>>     test, and remove mistaken and unfounded FIXME comment.
>
> Hello Stephen,
>
> this commit is introducing in my test-bed the following three failures
> in emacs-30: dabbrev-expand-test-other-buffer-4,
> dabbrev-expand-test-other-buffer-3, dabbrev-expand-after-killing-buffer.
> Could you please have a look?
>
>   Andrea

How does your testbed run the tests?  When I invoke `emacs-30 -batch -l
~/src/emacs/emacs-30/test/lisp/dabbrev-tests.el -f
ert-run-tests-batch-and-exit' from the shell, all tests pass.  Likewise,
when I start emacs-30 -Q, then `M-x load-file
~/src/emacs/emacs-30/test/lisp/dabbrev-tests.el', then kill all
non-hidden buffers except *scratch*, and then run `M-x ert', all tests
pass.  Since the batch run is also used by `make check', I've assumed
that, since the tests succeed (at least for me) in this mode, they are
valid regression tests.

In contrast, when I start emacs -Q, visit
~/src/emacs/emacs-30/test/lisp/dabbrev-tests.el, evaluate the buffer and
run `M-x ert', then I get five test failures:
dabbrev-expand-test-other-buffer-{1,2,3,4} and
dabbrev-expand-after-killing-buffer.  The latter fails because
dabbrev-expand finds an expansion in the test file buffer (but succeeds
when there is no such buffer, as in the previous run starting with -Q).
The four other tests fail possibly for the same reason; however, these
four tests fail in this way since being introduced in commit 8113b5c562
(see
https://lists.gnu.org/archive/html/bug-gnu-emacs/2024-10/msg01752.html),
so it puzzles me that you only see just two of those failures and only
after my latest commit.  (However, I did get the same three failures you
reported when I ran ert on the test file from commit bd8a6f70fb,
i.e. the one prior to the above commit, but only after uncommenting the
code lines that are commented out in that commit.  Could it be that you
accidentally tested using that previous commit with those code lines
uncommented?)

Steve Berman