nnir group marks don't propagate?

nnir group marks don't propagate?

[Eric Abrahamsen]

Am I hallucinating this, that marks set in an nnir search group used to
propagate to the "actual" server group? I could have sworn this worked,
as I think it used to be one of my major uses of nnir searches: to
"recover" old messages by ticking them or marking them unread.

Yet `gnus-summary-exit' pretty clearly checks for a quit-config, which
nnir/ephemeral groups have, and avoids `gnus-summary-update-info' if
it's present. And the code has been like that for yonks.

Am I really dreaming? Wasn't there some way to make search-group mark
changes propagate to the real backend?

Eric

Re: nnir group marks don't propagate?

[Andrew Cohen]

>>>>> "EA" == Eric Abrahamsen <eric@ericabrahamsen.net> writes:

    EA> Am I hallucinating this, that marks set in an nnir search group
    EA> used to propagate to the "actual" server group? I could have
    EA> sworn this worked, as I think it used to be one of my major uses
    EA> of nnir searches: to "recover" old messages by ticking them or
    EA> marking them unread.

Yes this should definitely work (and has worked for years and years). It
is still working in my setup (just checked) but I am using the
gnus-select feature branch.

The updating should NOT invoke gnus-summary-update-info; the nnir group info
list is not saved to the newsrc (unless you are using permanent search
groups). The trick is to update the info for the underlying group that
the marked messages are from, and this is done elsewhere. Exactly where,
err, I would have to go back and look at nnir.

Best,
Andy

Re: nnir group marks don't propagate?

[Eric Abrahamsen]

Andrew Cohen <cohen@bu.edu> writes:

>>>>>> "EA" == Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>     EA> Am I hallucinating this, that marks set in an nnir search group
>     EA> used to propagate to the "actual" server group? I could have
>     EA> sworn this worked, as I think it used to be one of my major uses
>     EA> of nnir searches: to "recover" old messages by ticking them or
>     EA> marking them unread.
>
> Yes this should definitely work (and has worked for years and years). It
> is still working in my setup (just checked) but I am using the
> gnus-select feature branch.
>
> The updating should NOT invoke gnus-summary-update-info; the nnir group info
> list is not saved to the newsrc (unless you are using permanent search
> groups). The trick is to update the info for the underlying group that
> the marked messages are from, and this is done elsewhere. Exactly where,
> err, I would have to go back and look at nnir.

Good to hear from you!

That's what I thought/hoped. I was sure I'd been using nnir like this
for years, and it seems to me that it was only in the past month or two
that it stopped working. Lars has returned from the dead and starting
fixing bugs, so that was my first suspicion, but it's really hard to
tell.

I had thought that `gnus-summary-update-info' did its thing because
`nnir-request-update-mark' would end up calling the same nnoo function
on whatever backend was *actually* in charge of the group in question.
But if that's not how it works, then I don't know where to look.
Potential culprits on master are 2ec41c41 and de296050, and maybe some
others in that neighborhood... I don't know how it's supposed to work.

Hey, merge your branch :)

Eric

Re: nnir group marks don't propagate?

[Eric Abrahamsen]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

> Andrew Cohen <cohen@bu.edu> writes:
>
>>>>>>> "EA" == Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>>
>>     EA> Am I hallucinating this, that marks set in an nnir search group
>>     EA> used to propagate to the "actual" server group? I could have
>>     EA> sworn this worked, as I think it used to be one of my major uses
>>     EA> of nnir searches: to "recover" old messages by ticking them or
>>     EA> marking them unread.
>>
>> Yes this should definitely work (and has worked for years and years). It
>> is still working in my setup (just checked) but I am using the
>> gnus-select feature branch.
>>
>> The updating should NOT invoke gnus-summary-update-info; the nnir group info
>> list is not saved to the newsrc (unless you are using permanent search
>> groups). The trick is to update the info for the underlying group that
>> the marked messages are from, and this is done elsewhere. Exactly where,
>> err, I would have to go back and look at nnir.
>
> Good to hear from you!
>
> That's what I thought/hoped. I was sure I'd been using nnir like this
> for years, and it seems to me that it was only in the past month or two
> that it stopped working. Lars has returned from the dead and starting
> fixing bugs, so that was my first suspicion, but it's really hard to
> tell.

I re-built an Emacs from around a year ago, and observed the same
problem, so I guess it's something in my setup, though I can't imagine
what (as usual).

If anyone else happens to see this, do let me know...

E