ert tests fail if racket-mode is installed

[bremner]

If the package is not installed, the tests pass. However, if the package is already installed,
two tests fail.

2 unexpected results:
   FAILED  racket-tests/indent-rkt
   FAILED  racket-tests/run

Here are the relevant backtraces

Test racket-tests/indent-rkt backtrace:
  (if (unwind-protect (setq value-52 (apply fn-50 args-51)) (setq form
  (let (form-description-54) (if (unwind-protect (setq value-52 (apply
  (let ((value-52 (quote ert-form-evaluation-aborted-53))) (let (form-
  (let ((fn-50 (function racket-tests/same-indent)) (args-51 (list "ex
  (lambda nil (let ((fn-50 (function racket-tests/same-indent)) (args-
  ert--run-test-internal([cl-struct-ert--test-execution-info [cl-struc
  ert-run-test([cl-struct-ert-test racket-tests/indent-rkt "Indentatio
  ert-run-or-rerun-test([cl-struct-ert--stats t [[cl-struct-ert-test r
  ert-run-tests(t #[385 "\306\307\"\203G\211\211G\310U\203\211@\20
  ert-run-tests-batch(nil)
  ert-run-tests-batch-and-exit()
  command-line-1(("-eval" "(progn (add-to-list (quote load-path) \"/ho
  command-line()
  normal-top-level()

Test racket-tests/run backtrace:
  (if (eq major-mode (quote racket-mode)) nil (error "Current buffer i
  racket--do-run(medium)
  racket-run()
  (let* ((racket--repl-command-connect-timeout (* 15 60)) (racket-comm
  (lambda nil (let* ((racket--repl-command-connect-timeout (* 15 60)) 
  ert--run-test-internal([cl-struct-ert--test-execution-info [cl-struc
  ert-run-test([cl-struct-ert-test racket-tests/run nil (lambda nil (l
  ert-run-or-rerun-test([cl-struct-ert--stats t [[cl-struct-ert-test r
  ert-run-tests(t #[385 "\306\307\"\203G\211\211G\310U\203\211@\20
  ert-run-tests-batch(nil)
  ert-run-tests-batch-and-exit()
  command-line-1(("-eval" "(progn (add-to-list (quote load-path) \"/ho
  command-line()
  normal-top-level()

[greghendershott]

1. How are you running the tests -- via make test?
2. What is the Emacs version?

[bremner]

Greg Hendershott <notifications@github.com> writes:

1. How are you running the tests -- via `make test`?

yes

2. What is the Emacs version?

25.1 by default; I ran "EMACSBIN=/usr/bin/emacs24 make test" and get the same results

[greghendershott]

Hmm. When you say "the package is already installed", do you mean using Emacs' package-install or something else?

[bremner]

Greg Hendershott <notifications@github.com> writes:

Hmm. When you say "the package is already installed", do you mean using Emacs' `package-install` or something else?

Using package-install, but into a system wide directory ╭─ zancas:~ ╰─% ls /usr/share/emacs25/site-lisp/elpa/racket-mode-0.4 channel.rkt racket-common.elc racket-mode.el cmds.rkt racket-complete.el racket-mode.elc defn.rkt racket-complete.elc racket-mode-pkg.el error.rkt racket-custom.el racket-mode-pkg.elc find-module-path-completions.rkt racket-custom.elc racket-profile.el fresh-line.rkt racket-edit.el racket-profile.elc gui.rkt racket-edit.elc racket-repl.el image.rkt racket-font-lock.el racket-repl.elc Install.log.gz racket-font-lock.elc racket-tests.el instrument.rkt racket-imenu.el racket-tests.elc keywords.rkt racket-imenu.elc racket-unicode-input-method.el logger.rkt racket-indent.el racket-unicode-input-method.elc mod.rkt racket-indent.elc racket-util.el racket-bug-report.el racket-keywords-and-builtins.el racket-util.elc racket-bug-report.elc racket-keywords-and-builtins.elc run.rkt racket-collection.el racket-make-doc.el scribble.rkt racket-collection.elc racket-make-doc.elc try-catch.rkt racket-common.el racket-mode-autoloads.el util.rkt

[bremner]

David Bremner <david@tethera.net> writes:

Greg Hendershott ***@***.***> writes: > Hmm. When you say "the package is already installed", do you mean using Emacs' `package-install` or something else? > Using package-install, but into a system wide directory

The problems seem to be caused by interactions with geiser. On the machine where the tests are failing I have (had) also installed geiser via package.el. At least for one of the tests, the problem seems to be related to geiser helpfully starting scheme-mode (now geiser) for rkt files. (add-to-list 'auto-mode-alist '("\\.rkt\\'" . scheme-mode)) At a guess, the fact that this only manifests as a problem when racket-mode is also installed probably has something to do with load order. If I think of a way to make the racket-mode tests more robust, I'll let you know.

[greghendershott]

Ah, interesting. I wonder what's a realistic scenario to test. I think most human users would install only one of these two packages -- or maybe, install both, but in their Emacs init file disambiguate mode they want as the default mode for .rkt files?

But to back up a step. It occurs to me now, need/ought you run packages' tests, at all?

I don't think that happens when people do the "normal" thing of using package-install?

(For example. I think some packages exclude test-related .el files from what MELPA distributes. Arguably it's wrong that the MELPA recipe for racket-mode is including them.)

[bremner]

Greg Hendershott <notifications@github.com> writes:

Ah, interesting. I wonder what's a realistic scenario to test. I _think_ most human users would install only one of these two packages -- or maybe, install both, but in their Emacs init file disambiguate mode they want as the default mode for `.rkt` files?

Both modes attempt to take ownership of .rkt files, but geiser is also useful for several other scheme variants.

But to back up a step. It occurs to me now, need/ought you run packages' tests, at all?

Yes, I'm working on packaging racket-mode for Debian and this kind of integration testing is part of the point. The underlying conflict in auto-mode-alist is something we'll have to resolve somehow.

I don't think that happens when people do the "normal" thing of using `package-install`?

That's true. I'm not really doing the "normal thing" in that sense.

(For example. I think some packages exclude test-related `.el` files from what MELPA distributes. Arguably it's wrong that the MELPA recipe for `racket-mode` is including them.)

I'm not downloading the files from melpa; I don't really know what the general melpa policy is about test suites. d

[greghendershott]

Thanks for explaining.

1. I'll think about this more.

2. In the meantime, my least-worst idea answer is that racket-mode expects that, if it is running, it's because the user wants to use it to handle .rkt files; otherwise it's not much use. :) The tests assume that. If the user's init file disambiguates which major mode is supposed to be set for .rkt files, and the answer is not racket-mode, then probably the correct thing to do is... not run the tests for the package that will not be used. Maybe the tests should check the .rkt mode and simply not run at all. idk

[greghendershott]

Or, maybe the tests that visit a .rkt file should explicitly call (racket-mode 1) -- similar to a user visiting the file, then manually doing a M-x racket-mode to set the mode.

The tests intentionally don't do that as an indirect way of making sure I set the auto-mode-alist and that it was effective. But, maybe that's "too strong" a test.

[bremner]

Greg Hendershott <notifications@github.com> writes:

Or, maybe the tests that visit a `.rkt` file should explicitly call `(racket-mode 1)` -- similar to a user visiting the file, then manually doing a <kbd>M-x</kbd> `racket-mode` to set the mode. The tests intentionally _don't_ do that as an indirect way of making sure I set the `auto-mode-alist` and that it was effective. But, maybe that's "too strong" a test.

I think the underlying issue is that calling "(package-initialize)" gives up control over the test environment. In a sense this is good, since we discovered a real world problem, but it makes the tests inherently unreprodicible. What do you think about setting 'package-load-list' to only load the packages needed for the tests? d