bug#51841: 27.2; erc-insert-marker has no value

[J.P.]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> I've now made `erc-mode' noninteractive, because it's not something
> users should call themselves (as you've found out).

Nice. (Of course I knew `define-derived-mode' can take an :interactive
keyword. Honest. I swear.)

                                 * * *

Rostislav Svoboda <rostislav.svoboda@gmail.com> writes:

> BTW I tried to reproduce the problem and I realized the bug gets
> triggered only if I run the `M-x erc-mode` in the "main" erc buffer.
>
> [...]
>
> The bug doesn't get triggered when running the `M-x erc-mode` in a
> channel-buffer.

I have a theory that might account for the disparity here [1].

> I started erc with:
>     erc-fill-function 'erc-fill-static
>     erc-fill-static-center 20
> and realized that empty space of 20 chars is too much. So I ran `(setq
> erc-fill-static-center 15)` only to see no change.

Hm, does "no change" refer to future output? If so, then we may have an
actual bug because that definitely should be affected [2], right?

> (That doesn't activate the change either, but that's another story,
> though :)

But by "activate" (and also from your other comments), I'm starting to
think you mean not only for subsequent output but for all prior output
as well. What I mean is, are you expecting the buffer to be refilled
according to the updated value? If so, ERC doesn't support that (though
I think Circe might). If that's what you're after, that would make a
good feature request. We could even change the title of this to
something like "Support dynamic refilling of ERC buffers". (Or just open
a new bug.)


[1] FTR, I think what was going on was that as soon as you issued the
    command in a live channel buffer, ERC considered it closed (because
    it appeared to have suddenly lost its process). So as soon as a new
    message for that channel arrived, ERC created a replacement buffer,
    which was what you probably observed as behaving normally. However,
    if you had tried switching to the CLOSED buffer, it might've still
    shown the bug had you typed something and tried to send it, I think.

    
https://jpneverwas.gitlab.io/-/erc-tools/-/jobs/1782289657/artifacts/logs/51841/remode/replay.html

[2] (Spoiler alert: could not reproduce)

    
https://jpneverwas.gitlab.io/-/erc-tools/-/jobs/1782289657/artifacts/logs/51841/dynamic-fill/replay.html