-
Notifications
You must be signed in to change notification settings - Fork 41
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Consider the GitHub "Stale" application for backdrop-issues #4172
Comments
I've seen this (or similar bots) used in other issue queues, and I can see the benefit. Though we'd want to avoid a situation like:
So yeah, creating a |
A few random thoughts about this:
I could support using a tool like this if we could do the following:
|
@stpaultim Thanks for the feedback. It looks like you can set a label that's applied when an issue's marked as stale. That means there will be ways you can see stale issues, sort by most recently updated, etc. That might address some if your concerns. Here's some better documentation about what's possible with the bot: https://github.com/probot/stale/blob/master/README.md |
I knew it; I'm being replaced by a robot! 😄 |
@stpaultim some random answers to your random thoughts... 😉
Maybe yes, maybe no. If an issue queue fills up, people often miss an already existing one and create a new issue anyway.
It seems, that's exactly how this robot works. It gives a warning (comment = mail), adds a (configurable) label and after a configurable time span with no activity the issue gets closed. Another thought: if issues get really old, they need re-evaluation anyway. It might be, the problem was already solved by another issue. @BWPanda Noooooo.... no robot could ever replace you! 😂 |
I had similar thoughts as @stpaultim, and I see that the concerns can be addressed with a reasonable configuration of the bot. |
This was discussed in some depth during the DEV meeting today (https://youtu.be/ot7ug0NvdFY) There were some fairly strong objections to automatically closing issues. There is a strong concern about how the impact that might have on the people who took the time to report issues, to see them closed at a later date automatically. There is a sense that an issue that is unresolved should just stay open until it's resolved and that it's better to have a human triage these old issues and decide what should be closed. There was interest in adding tools that would make it easier for humans to triage and close old issues. We primarily discussed two options.
@jenlampton expressed some concerns about too many labels and encouraged us to create a useful link to a predefined filter that would help users triage these old issues. I was just looking for a way to filter by a FYI - there are currently 186 open issues that have not been updated in the last 3 years. We could work our way through them and then change the filter. But, for this to work, we'd want to update each of these issues so that we don't end up retriaging them over and over. We could look at each of these issues and either close them or post a short comment saying that that it is still a relevant issue. If we were do this, we'd in effect be saying that our goal is to NOT have any open issues that have not been commented on for more than 3 years (or some other time frame) and then either close or comment on each one until we reach our goal. The downside would be that many of these issues would float to the front of the "updated queue" without any real new information. NOTE: I don't feel strongly about this, but would be willing to help triage old issues from time to time. |
I personally disagree with the objections to this proposal. I think a carefully and respectfully worded message would suffice just as well as manually reviewing old issues. If the original author is still around/interested, they can reply to keep the issue open (same as if I manually ask them if it's still relevant). And if they don't care anymore they can ignore it and it'll be closed (which from experience I've seen is the case with the majority of old issues). I don't see any benefit to the manual process vs. an automated one, only more work, time and effort that could be better spent on other issues. |
I second that. Of course, we don't want to be rude and close issues too early or in a disrespectful way, but I'm pretty sure, we're able to use that application without giving people a feeling of "a bot killed my contribution". Please note: we're talking about times of inactivity of one or even more years, at least several months. If no one showed any interest on it for such a long time, and it's none of our meta tasks, this issue is obsolete for sure. At least to my understanding. Another concern was: Issues should only get closed, if the problem is really fixed. Anyway, this is an interesting discussion. I'm curious, what the final decision will be. |
I'm all for using the bot. It's configurable so we can make sure that we're transparent and that we're considerate. And even if we decide not to close the issues, we could use a different action to just label them as stale to help people triage. |
Also, as discussed in Gitter, GitHub now has native support (currently in beta) for actions such as Stale: https://github.com/actions/stale |
Should this issue be marked as 'stale'? 😆 But seriously, do we continue to push for this or not? |
I'm fine with this, so long as we don't auto-close:
|
Yeah, I think we should trial this by adding a new |
I have just setup this on the Auto Entity Label repo as a test: If we want to test it here, we need to decide:
For the wording, I think something like this:
|
Here's a PR: #4757 |
One year later and after seeing myself fixing several pretty old issues, I'm not so convinced of the "stale" concept anymore. 😆 Anyway, I like @BWPanda's suggestion to do some cautious testing with only adding a label (no autoclose) and see how this affects our workflows. The name could be more precise.. "Mark old issues as stale" - I don't think it's the age that causes a stale label. How about "Mark stale issues"? |
I'm fine with auto tagging issues that are inactive. In fact, if we are auto tagging old issues, I prefer the term "inactive" or something like that. However, I'm fine with "stale." If we do this, I would say that anyone can remove the "stale" or "inactive" tag at any time. I would not tag an issue as inactive for at least a year (I would prefer 2 years). I find it pretty common that issues 6-12 months old suddenly get very active and issues that have a lot of interest, sometimes go quiet for 6 months or more without becoming less important (just less urgent). I don't really see any benefit to auto closing old issues, but am happy to leave that discussion on the table for later. |
I'm against auto-closing. There's nothing more sad then searching for a bug
you are experiencing and finding a github issue that has been auto-closed
with no human interaction other than the original report. It's a
contribution killer. Why would that first person ever contribute anything
again?
What does it mean for an issue to be stale? We can already reverse-sort
issues by recent activity.
I feel similarly concerned for our contributors when adding stale label
automatically. We already have a "cannot reproduce" label that people can
add when appropriate.
…On Thu, Nov 12, 2020, 11:46 PM indigoxela ***@***.***> wrote:
One year later and after seeing myself fixing several pretty old issues,
I'm not so convinced of the "stale" concept anymore. 😆
Anyway, I like @BWPanda <https://github.com/BWPanda>'s suggestion to do
some cautious testing with only adding a label (no autoclose) and see how
this affects our workflows.
The name could be more precise.. "Mark old issues as stale" - I don't
think it's the age that causes a stale label. How about "Mark stale issues"?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#4172 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AADBER3Z7UO2TIWYLHJOU5LSPTP47ANCNFSM4JGV5GPQ>
.
|
Yes but how often do we actually do that? The benefits here are:
|
If people aren't using the tools we already have, I don't see how adding a
duplicate one that's slightly harder to use will make them use it more. A
reverse sort on the queue is already only one click.
We already use a lot of labels and n the core queue, I don't really want to
add another one that indicates something we already know.
But most importantly, I am concerned about demoralizing contributors.
If the primary benefit is the posting of a comment, why don't we automate
that part, but with some more helpful and friendly language?
…On Fri, Nov 13, 2020, 12:39 PM Peter Anderson ***@***.***> wrote:
What does it mean for an issue to be stale? We can already reverse-sort
issues by recent activity.
Yes but how often do we actually *do* that? The benefits here are:
- Old issues are *automatically* tagged (and therefore slightly easier
to find - means you can filter issues by 'stale' label *and* sort by
date (asc or desc))
- But primarily: the bot *automatically* posts a comment saying the
issue is 'stale' which prompts a response from the participants (who get
notified)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#4172 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AADBERYTBWM6O2BE23Q3273SPWKRNANCNFSM4JGV5GPQ>
.
|
I have to admit that that's what I do as well when I feel like revisiting the oldest issues in our queue. |
I was asking for suggestions for the wording of the message, so please feel free to share ideas. |
It seems that we all agree that auto-closing isn't what we want.
That's where I see the benefit of it. |
I'm losing interest in this issue. There doesn't seem to be much support for it. Maybe we should close for now and possibly re-visit later...? |
Agreed. |
Issues queues tend to fill up with stale issues over time. GitHub offers a free application for this use case.
https://github.com/marketplace/stale
It also allows to define exemptLabels which prevent issues from auto-closing.
I wonder, if this could be useful to keep backdrop/backdrop-issues clean(er), so I'd like to start a discussion here.
Personally, I never used that application, but the description sounds appealing to me.
The text was updated successfully, but these errors were encountered: