bug#62694: 30.0.50; eglot-tests fails with recent pylsp

[Michael Albinus <michael.albinus@gmx.de>]

João Távora <joaotavora@gmail.com> writes:

Hi João,

>> I disagree. Running regression tests towards bleeding edge development
>> version of pylsp is not the intention. You cannot expect, that everybody
>> running Emacs tests has installed pylsp like this. But she could have
>> installed the Debian pylsp package.
>
> Sure, or she has a 'pylsp' that prints the complete works of
> Shakespeare. But these tests are designed for the number one
> recommended pylsp installation method. It's AFAIK not the
> "bleeding edge" (which would be a Git installation). And tests
> have been running fine with that installation method for a
> number of months now, maybe even years.

I don't doubt. But you cannot expect that everybody uses "the number one
recommended pylsp installation method".

>> I have used the most recent pylsp package from Debian. If it doesn't
>> cooperate with eglot, we have a problem.
>
> We? I'd say Debian has, right?  I don't have anything against it,
> quite the contrary, but I don't have Debian, and I don't program
> against it (that's also a reason why I can't debug this right now).

No, "we" (as the Emacs community) have a problem. Everybody who has
installed the Debian pylsp package, and who runs Emacs' "make check",
will see the errors. And these are not counted as pylsp errors, or
eglot.el errors; they are counted as Emacs errors. We shouldn't show
such a bad performance.

If Debian's pylsp does not cooperate, add a check in eglot-tests.el that
it is skipped.

Testing Debian and/or Emba is pretty easy now. I have just pushed a
patch to master, extending the file admin/notes/emba. There is a new
section "Running Emba tests locally", which you could apply. Use the
build target emcs-eglot instead of emacs-inotify, and run finally "make
-C test eglot-tests", and you'll see what's up.

> I can probably switch to tests to use some other server, maybe
> one whose Debian package is more well-behaved, but this is
> hardly a priority for me.

A priority is that the tests shouldn't fail. That's for all us
developers.

> What/whose problem or scenario are you trying to solve?
> Who is this hypothetical she-user and what is she trying
> to do?  Develop Emacs, develop Eglot, run eglot tests, use
> Eglot with python? All/some of the above?

See above. A simple "make check" could already fail, with an improper
pylsp installed.

> What is preventing you or her from installing this external
> tool using its recommended installation method?  Does Debian's
> python not carry its 'pip' package manager? Why do you want to
> install it, and why do you want to install it specifically like this?

Again, I'm not speaking about eglot users. They shall know what to
do. But eglot-tests could fail for everybody who has installed a pylsp
package, for example from Debian, w/o even being interested in eglot.

>> I recommend that you investigate why the tests fail. And if this sounds
>> too much, at least the sanity checks shall be improved. (skip-unless
>> (executable-find "pylsp")) doesn't seem to be sufficient then.
>
> I don't have Debian.  We can overhaul the sanity checks, but I don't
> immediately see how.  Or why.  So it's hardly a priority.

See above. With the instructions I have added to admin/notes/emba, it
should be simple.

> João

Best regards, Michael.