Re: Eglot tests fail on emba

João Távora <[email protected]> writes:

Hi João,

>     Use of file notification itself doesn't seem to break eglot, as
>     much as I understand. There are two questions:
>
>     - Why does file notification speaks about creation of .#main.rs
>     now?  This file wasn't mentioned before, and I don't understand
>     where it comes from.
>
> I don't know. All I can say as I said before is that Eglot and its
> immediate neighbors haven't changed.
>
>     - Which process sentinel is it that the broken test speaks abouut?
>
>     --8<---------------cut here---------------start------------->8---
>     error in process sentinel: Process EGLOT
>     (on-type-formatting-project/(rust-mode)) not running: finished
>     --8<---------------cut here---------------end--------------->8---
>
>       file notification doesn't use processes.
>
>     I have no idea where to get the answer, and I believe somebody
>     with eglot knowledge should be involved.
>
> As I said, I'm not with a laptop for two weeks. 
>
> I think that's an irresponsible stance that unjustly holds hostage the
> users of Emacs master. What would you say if I put something in that
> broke TRAMP in some subtle way and refused to revert it until it my
> understanding of the issue met some arbitrarily high bar?
>
> You know you broke something. You don't know exactly what you broke,
> and you definitely didn't do it on purpose. And that happens, it's a
> part of dev life. Happens to me all the time. But it's pretty clear
> your actions, not anyone else's, are responsible for this breakage.
> Act in accordance with this established fact. 

Perhaps there is a bug in my change. But maybe, it is just a bug lying
around not in file notification, which was uncovered by the slighly
changed race conditions in file notification? We don't know this.

> There is a much of high-minded talk in this list about strict backward
> compatibility and how valuable it is and what it means, so let's act
> accordingly.
>
> So imo opinion it's fine to give it some time while you analyse, and
> to enlist whoever you think can help (I can't as I've explained) but
> if you find yourself blocked, I think the responsible thing to do is
> to put your work to the side until you the reasons that led to this
> breakage are known. I promise to help when I get back in September. 
>
> Also, have you tried my suggestion to micro-bisect your change, the
> commit I pointed to, and which we know broke things? I think it could
> yield insight!

Running eglot-tests interactively doesn't show a problem: all tests
pass. And yes, I have done this in the VM created with emba environment.

So there is no emergency: we can wait until you return, and have time
for checks.

> João

Best regards, Michael.