llvm dev - Oct 2021 - Proposal: introduce dependency on abseil when building benchmarks

On Thu, Oct 7, 2021 at 9:12 AM Renato Golin <rengolin at gmail.com> wrote:
> On Thu, 7 Oct 2021 at 17:01, Mircea Trofin <mtrofin at google.com>
wrote:
>
>> Unless people that actually build and run benchmarks have agreed with
>>> your proposal, you should not merge a clear breaking change.
>>>
>> One of the goals of this thread was identifying who those folks may be.
>>
>
> And yet, you propose to follow through if no one objected. Which is never
> a good idea for breaking changes.
>
To be clear, the fact they are breaking changes is what we're trying to
determine. It was my intention to spur attention to the thread (given the
silence), and providing a timeline can help.
>
>
It's really easy to miss an email like this and it's just out of luck
that
>>> I didn't.
>>>
>> Is there a more appropriate channel of communication where owners could
>> be identified?
>>
>
> No, this is the right place.
>
> But you either wait for people to find this thread (however long it
> takes), or you actively search for them (as I outlined before) and include
> them in the conversation, for example, CC'ing them in the thread.
>
Yup, but they need to be identified first (i.e. a bit of a catch 22 if no
one replies)
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20211007/d5831079/attachment.html>

There's probably no better mailing list, etc. It's an unfortunate
situation
that active developers don't/can't keep up with llvm-dev (it's a
lot, even
without commits lists, etc) such that it generally isn't sufficient to
email a non-targeted email to llvm-dev and expect to find the interested
parties in some particular subsystem/feature area/etc. Generally if you
don't already know at least a representative subset of interested parties
you may need to go looking for them similar to finding appropriate code
reviewers - check commit histories in the code in question, commit
histories in buildbot configurations that might be using this feature (in
this case), and maybe then buildbot owners that enable the configuration.
If you can find at least a couple of interested parties in such a search
and nothing else turns up that may be adequate to provide a variety of
perspectives/sufficient consensus to make a decision and/or they might be
able to point to others they know of who are working in that area.

On Thu, Oct 7, 2021 at 9:20 AM Mircea Trofin via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
>
>
> On Thu, Oct 7, 2021 at 9:12 AM Renato Golin <rengolin at gmail.com>
wrote:
>
>> On Thu, 7 Oct 2021 at 17:01, Mircea Trofin <mtrofin at
google.com> wrote:
>>
>>> Unless people that actually build and run benchmarks have agreed
with
>>>> your proposal, you should not merge a clear breaking change.
>>>>
>>> One of the goals of this thread was identifying who those folks may
be.
>>>
>>
>> And yet, you propose to follow through if no one objected. Which is
never
>> a good idea for breaking changes.
>>
> To be clear, the fact they are breaking changes is what we're trying to
> determine. It was my intention to spur attention to the thread (given the
> silence), and providing a timeline can help.
>
>>
>>
> It's really easy to miss an email like this and it's just out of
luck that
>>>> I didn't.
>>>>
>>> Is there a more appropriate channel of communication where owners
could
>>> be identified?
>>>
>>
>> No, this is the right place.
>>
>> But you either wait for people to find this thread (however long it
>> takes), or you actively search for them (as I outlined before) and
include
>> them in the conversation, for example, CC'ing them in the thread.
>>
> Yup, but they need to be identified first (i.e. a bit of a catch 22 if no
> one replies)
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20211007/3cbe9864/attachment.html>

Ok, it looks like we're talking past each other, so I'll be extra clear.
This is not meant to be rude, just making sure we're in the same page.

On Thu, 7 Oct 2021 at 17:20, Mircea Trofin <mtrofin at google.com> wrote:
> And yet, you propose to follow through if no one objected. Which is never
>> a good idea for breaking changes.
>>
> To be clear, the fact they are breaking changes is what we're trying to
> determine. It was my intention to spur attention to the thread (given the
> silence), and providing a timeline can help.
>
There are many ways to drive attention to a thread that hasn't received the
due attention, for example:
* "Politely ping" the thread, by saying "hey, has anyone seen
this, I need
some feedback" or such.
* Reply to the thread including more people and making it clear that you're
doing so by saying: "Adding more people to broaden the search for
reply".
* Chase other channels, like IRC, Disco{rd|urse}, other tools that you know
are people that can help.

Your "gentle reminder" isn't good because it assumed(*) that:
* everyone that should have read your email did read the email
* your change is obviously-good &| non-breaking
* silence with consensus

None of which is true.

(*) Even if you personally didn't assume, the reply itself had the
consequence (merge EOW) as if it was true.

Yup, but they need to be identified first (i.e. a bit of a catch 22 if
no
> one replies)
>
Both LLVM's mailing list and IRC channels are high-traffic, meaning it's
really easy to get lost in the noise. If you don't get a reply in a few
days, it's very likely that no one saw it.

People also work no different time zones, have different week schedules,
holidays, sickness, etc. Some people have alerts for emails, others read a
digest at the end of the day. Some people work on the project full time,
others on their spare time.

Most people, however, are more responsive if you copy/@ them directly than
if you just post a generic thread somewhere. But how to find those people?

First, old timers know more people than new contributors, so asking them is
a starting point.

But most of them will say similar things:
* Look for the code owners, buildbot maintainers
* Search the mailing list, phabricator or bugzilla for past topics and copy
the people involved
* Use a different communication channel (ex. IRC) to ask around

Once enough people are involved, they themselves know who else to copy, and
the process becomes a lot easier.

Now, where to look for people?
* Code owners are listed in the file: llvm-project/llvm/CODE_OWNERS.txt
* Infrastructure maintainers are listed in:
llvm-project/llvm/RELEASE_TESTERS.TXT and
https://lab.llvm.org/buildbot/#/workers (among others)
* Mailing list and bugzilla searches aren't very effective, but help spot a
few people.

Once you get enough people, just copy them on the initial RFC, and it will
have a much higher probability of reaching the right people.

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20211007/cfe80653/attachment.html>