Stop bugzilla bot from marking issues as "RESOLVED ARCHIVED".

[Augie Fackler]

On Wed, Feb 27, 2019 at 03:31:11AM +0530, Faheem Mitha wrote:
>
> Hi folks,
>
> See below.
>
> On Tue, 29 Jan 2019, Pulkit Goyal wrote:
>

[snip]

> Frankly, if you want to keep old bugs around, I can hardly think of a
> better way than by closing the bug reports periodically for no
> reason.
>

No reason is in fact a reasoned stance. See below.

> Who thought this,

It was discussed at a sprint years ago, and the room was in general
agreement that it seemed better than the (at the time) status quo of
an insurmountable pile of unloved bugs. It's an outgrowth of a process
used at Facebook, and my understanding is that it's effective there.

> frankly deranged,

Please dial back the attitude. I understand that not everyone agrees
with our bug-tracking setup (heck, I'm not sure I do), but
name-calling is unwelcome here.

> practice was a good idea, anyway?

Without the bot, the bug backlog grows without bound. Nobody actually
checks to see if old bugs are fixed, and eventually the bugtracker
becomes a disaster and not useful. I've, broadly, seen four
disciplines around OSS project bugtrackers:

1) There's a bug tracker and some people spend most of their time on
the project closing bugs (not coding: just cleaning up ancient bugs).

2) "Nothing": the bug backlog grows roughly without bound. Eventually
developers stop looking for bugs to fix.

3) A robot auto-closes old bugs, with the expectation that humans
reopen them if someone still cares. (This is where we are. It's also a
semi-intentional CADT approach: https://www.jwz.org/doc/cadt.html)

4) Have no bugtracker (famously Linux does this, I think Git does too).

I don't really feel strongly (I've basically given up on the
bugtracker as a way to find interesting problems to fix already), so
if the consensus is that we'd rather be in state (2) than state (3)
that's fine with me, I guess. Note that I don't think it's any better:
it just means that you can shout into a void and not know it's such,
as opposed to now where at least you get some kind of notification
that your bug is unloved.

Other options that come to mind: we could make an "unloved" bug tag
that bugs could automatically move to (instead of closing, they move
to the unloved state). We could lengthen the amount of time bugs sit
open before the bot closes them. Sadly, I can't expect that we'll ever
hit state (1), where bugs regularly get groomed and fixed or
explicitly wontfixed - there's just not enough motivation on the part
of those developing hg to do something about it, as far as I can tell.

BTW, if you want a long (but entertaining, and IMO informative) blog
post on the topic, https://apenwarr.ca/log/?m=201712 has a lot to say
on the topic of bug backlogs. If you can figure out how to encourage
(or are volunterring to do) the bug triage Avery describes, I'm
definitely interested.


Thanks,
Augie

>
>
> Regards from an old user. First reported Mercurial bug circa 2005.
>
> Faheem Mitha
> _______________________________________________
> Mercurial-devel mailing list
> Mercurial-devel at mercurial-scm.org
> https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel