I see your idea here, Igal, but I don't think it's necessary.
Developers usually have pretty good tools to see where a bug came from (for example, we have a tool called "git bisect"[1] that allows us to analyze not just which release the bug was introduced in, but a specific commit to "blame" for the change. We use it rarely, though, because usually a ticket with good information about the bug is enough.
If you report a bug telling us what browser you've used, what you did, precisely (clicked on X, opened Y, looked at feature Z, typed A, etc etc) and where (what page, what wiki, etc) then we usually are pretty good at pinpointing the issue, and if we're not, we tend to know what to look for and ask the user for more information.
When we run into trouble understanding a bug, it is rarely due to knowing whether it is the new release or not. It's usually other factors that are (usually) a lot more important for us to know, like any gadgets that a user has, or some browser extension, or some edge case we need to figure out how to reach. From my experience (and I think, from the comments on the thread, from other devs experiences too) the "which release this came from" is less of an issue.
I think having another wiki to maintain versions for (there's still a cost for this in terms of maintenance and time, etc) is unnecessary for these points you raise...
Did you encounter specific times where a developer couldn't figure out a bug because they didn't know whether it was old or new? It might be that the issue has a different cause, and we can tackle this challenge differently for a good solution.
[1] https://git-scm.com/docs/git-bisect
On Fri, Sep 22, 2017 at 5:29 AM, יגאל חיטרון khitron@gmail.com wrote:
The purpose is to say to developers if it's a new problem, or isn't. I can think about six benefits:
- It can save them the time for checking this.
- It can be made better by task filer than by developer, because the first
knows better the problem. 3) It can save the time needed to them or other to find duplicate task, because there are less tasks to check, only from the last week. 4) There are many cases when the filer have no way to give an instruction to reproduce the bug, does not succeed to create such an instruction, or the developer does not succeed to reproduce it. If the developer will know that the bug is new, it will be much easier to they to recognize the problem, or find the right way for reproduction. 5) Groups 0-1-2 are done (also) exactly for this. New wiki will just give 5 times more time to see the bugs, which not always come in the first day only. 6) If there is some new bug that wasn't found on groups 0-1, but was found, say, by enwiki user - and most of not tech users is there - they have not another day to compare, as previous groups do. About logs - I don't talk about extremally techs users, almost developers, that can do all the research work by themselves, but about the most of them. Igal
On Sep 22, 2017 11:34, "Eran Rosenthal" eranroz89@gmail.com wrote:
If you don't have access to old mediawiki version (whether it is group2, your own wiki or test3wiki suggested above), and suspects there is a regression of something that was working in the past, it is useful to indicate it in the bug description, and the maintainer of that feature can check it out (either by indications from git log with relevant commit messages, or by restoring to older code)
You can also go over the commits log using web interface: https://github.com/wikimedia/mediawiki/commits/master and as Niharika said there are a lot of new changes :) While this is static (compared to running mediawiki instance) it as its
own
advantages: You can find out who is the owner (blame), related bugs (usually "Bug: X" in commit message) etc
On Fri, Sep 22, 2017 at 8:27 AM, Niharika Kohli nkohli@wikimedia.org wrote:
When filing a phab task with some new bug, you always want to know - is it really new, or I just did not pay
attention
to it before?
What's the purpose of this information? If it's a bug, new or not, a
ticket
needs to be filed.
And when I do know it's a new bug, I can open both versions
in the same time, and compare the behaviour for this bug. And also,
compare
the console results - what exactly changed in html, in css, in js
commands
reactions.
I agree that information will save some developer time but at the same
time
this information is not so easy to gather. This is helpful when the
users
have some working knowledge of how developer tools work and how to
compare
file changes. Usually in each version there are a lot of new changes.
Often
it's not easy for developers even to find out what could be causing the bug.
I can easily imagine such a wiki quickly falling into disuse.
On Thu, Sep 21, 2017 at 7:29 PM, יגאל חיטרון khitron@gmail.com
wrote:
It can work. But another Monday. I mean, if Tue-Wed-Thu there is a deployment of version ....5, a day before, Mon there is a deployment
of
version ....4, so starting from tomorrow, group 0 will get a way to
see
both version, exactly from the beginning, but not until the end, for
6
days, group 1 for 5 days, and group 2 for 4 days. And from Monday to
the
deployment, 1-2-3 days, there will not be use of this. I'll be very
glad
if
it will be decided to do this, and if so, it will be a good thing to
add
to
the text of how to report a bug in phabricator help, something about,
you
can check if it is a regression, the last version "falt", by
comparing
with
this new wiki. I can thing about many dozens of tasks I wrote and
read
where this information could be useful, if added at the first place.
Hope
you decide this indeed. Thank you very much, Igal
On Sep 22, 2017 05:17, "Chad" innocentkiller@gmail.com wrote:
No non-emergency deployments on Fridays, Saturdays or Sundays. Monday could work.
-Chad
On Thu, Sep 21, 2017 at 7:15 PM יגאל חיטרון <khitron@gmail.com
wrote:
I glad you say so. What about Friday? Igal
On Sep 22, 2017 05:07, "Chad" innocentkiller@gmail.com wrote:
> It wouldn't be hard to do at all, technically. I imagine it'd
be
something > like a test3wiki. > > Main thing to know is when do we cycle off of the old version?
When
the
> version goes out on Tuesdays? That day's already pretty loaded
for
software > moving about... > > -Chad > > On Thu, Sep 21, 2017 at 1:25 PM Brian Wolff <bawolff@gmail.com
wrote:
> > > Making your case here is probably best. The release
engineering
team
are > > the people you probably have to convince, although of course
anyone
could > > potentially create such a wiki, in an unofficial way. > > > > Keep in mind that keeping an older version of the software
running
does
> > introduce a maintinance burden, so you will probably have to
convince
> > people that it would be regularly useful and not just useful
this
one
> time. > > > > -- > > bawolff > > > > On Thursday, September 21, 2017, יגאל חיטרון <
khitron@gmail.com>
wrote: > > > Thank you. Sorry to hear this. Is there some place I can
suggest
this
> and > > > explain why do I think it can be very helpful? > > > Igal > > > > > > On Sep 21, 2017 22:12, "Brian Wolff" bawolff@gmail.com
wrote:
> > > > > >> On Thursday, September 21, 2017, יגאל חיטרון < khitron@post.bgu.ac.il> > > >> wrote: > > >> > Hi. Sometimes after the week deployment I need to
compare
the
new
> > version > > >> > with the previous one, in some aspect. Is there a test
wiki
that
> > always > > >> has > > >> > one version before the current? > > >> > Thank you. > > >> > Igal (User:IKhitron) > > >> > _______________________________________________ > > >> > Wikitech-l mailing list > > >> > Wikitech-l@lists.wikimedia.org > > >> > https://lists.wikimedia.org/mailman/listinfo/wikitech-l > > >> > > >> No there is not. You can of course download old versions
of
the
> software > > >> and setup your own wiki but that is a lot of effort. > > >> > > >> -- > > >> bawolff > > >> _______________________________________________ > > >> Wikitech-l mailing list > > >> Wikitech-l@lists.wikimedia.org > > >> https://lists.wikimedia.org/mailman/listinfo/wikitech-l > > > _______________________________________________ > > > Wikitech-l mailing list > > > Wikitech-l@lists.wikimedia.org > > > https://lists.wikimedia.org/mailman/listinfo/wikitech-l > > _______________________________________________ > > Wikitech-l mailing list > > Wikitech-l@lists.wikimedia.org > > https://lists.wikimedia.org/mailman/listinfo/wikitech-l > _______________________________________________ > Wikitech-l mailing list > Wikitech-l@lists.wikimedia.org > https://lists.wikimedia.org/mailman/listinfo/wikitech-l _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
-- Niharika Software Engineer Community Tech Wikimedia Foundation _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l