bug#59742: server-tests failures

bug#59742: server-tests failures

[Mattias Engdegård]

The server-tests still fail for an apparent variety of reasons. The behaviour differs depending how the test is run: from a terminal, from ´M-x compile`, interactively via `M-x ert`, Emacs started with or without `-nw`, and so on.

At least the batch-mode tests would cease failing if we just applied the following:

 (defconst server-tests/can-create-frames-p
-  (not (memq system-type '(windows-nt ms-dos)))
+  (and window-system (not (memq system-type '(windows-nt ms-dos))))

Any reason not to?

Even so, there is still a failure if run interactively (M-x ert), and a created frame isn't removed properly.

bug#59742: server-tests failures

[Jim Porter]

On 12/1/2022 7:01 AM, Mattias Engdegård wrote:
> At least the batch-mode tests would cease failing if we just applied the following:
> 
>   (defconst server-tests/can-create-frames-p
> -  (not (memq system-type '(windows-nt ms-dos)))
> +  (and window-system (not (memq system-type '(windows-nt ms-dos))))
> 
> Any reason not to?

These tests pass on GNU/Linux when 'window-system' is nil, so I think 
we'd need a different way to check things.

> Even so, there is still a failure if run interactively (M-x ert), and a created frame isn't removed properly.

I'll see if I can improve the tests when run interactively, though I'm 
not sure that will affect the first part of this bug.

bug#59742: server-tests failures

[Mattias Engdegård]

1 dec. 2022 kl. 18.24 skrev Jim Porter <jporterbugs@gmail.com>:

> These tests pass on GNU/Linux when 'window-system' is nil, so I think we'd need a different way to check things.

They evidently don't pass on the (Linux) CI server either. No wonder, if they rely on the ability to create new frames.

bug#59742: server-tests failures

[Mattias Engdegård]

I pushed (to emacs-29) an experimental change to fix one of the test that didn't clean up its frame, and to avoid running frame-creating tests when TERM=dumb which is what M-x compile uses.

Please give it a try. Let's see how the CI takes it.

bug#59742: server-tests failures

[Mattias Engdegård]

> Let's see how the CI takes it.

Looks like it's happy now. Closing now, but do tell if you have misgivings.

bug#59742: server-tests failures

[Jim Porter]

On 12/3/2022 7:20 AM, Mattias Engdegård wrote:
>> Let's see how the CI takes it.
> 
> Looks like it's happy now. Closing now, but do tell if you have misgivings.

Thanks. I finally had a chance to look into this myself. I pushed one 
tiny fix (3785fe52e4692ffef14c0a1e50361c22d66fabe8) so that the tests 
pass on GNU/Linux when run from a terminal.

bug#59742: server-tests failures

[Mattias Engdegård]

10 dec. 2022 kl. 08.02 skrev Jim Porter <jporterbugs@gmail.com>:

> I pushed one tiny fix (3785fe52e4692ffef14c0a1e50361c22d66fabe8) so that the tests pass on GNU/Linux when run from a terminal.

Looks fine; now back-ported to emacs-29.
(These days we mostly prefer `dolist` to `mapc`, but there is no reason to change your code!)