Hi everyone, Thank you all for the feedback. I completely understand the concerns about test suite maintenance and the risk of confusing arbitrary current behavior with the intended design.
To address Ihor's point about explicitly marking this test, I would like to propose a few ways we could handle this. 1. Naming Convention We could rename the test to explicitly include the word "characterization" (e.g. test-org-colview/characterization-insert-columns-keyword). Running it: BTEST_RE='characterization' test-dirty. Within Emacs: M-x ert RET characterization RET. 2. ERT Tags We could add a specific metadata tag, such as :tags '(characterization). Rnning it: M-x ert RET (tag characterization) RET or even (not (tag characterization)) to exclude them. However, our Makefile does not have a convenient way to filter by ERT tags. Running it from the CLI requires bypassing make entirely with a verbose batch command like: emacs -Q --batch ... --eval "(ert-run-tests-batch-and-exit '(tag characterization))". But I haven't seen tags in our tests sutis. 3. Docstring and "Pinning" Comments We could add a clear disclaimer in the docstring and a comment right above the assertion, explicitly stating that this is not a design spec. For example: "Record the current insertion position of a new #+COLUMNS keyword. NOTE: This pins a known arbitrary behavior to prevent regressions. It does NOT represent the intended design spec. See mailing list thread <...>" Personally, I think combining all three would make the intent crystal clear. Which of these approaches (or what combination) would you prefer? Once we agree on the format, I will update the patch. Best, -- Slawomir Grochowski
