Is the LOG= line still useful?

[Adam Klein]

Back in the day, when we were still on SVN and all V8 releases were blobs of merged changes, the LOG=Y served a good purpose, giving those SVN commits on the release branch human-readable change descriptions.

Now that our releases are merge commits on branches with full CL descriptions, and moreover the release commits contain a link to a full list of changes, I'm not sure the LOG stuff is buying us anything other than nags from the presubmit script.

How would folks feel about dropping it?

- Adam

[Michael Hablich]

My impression is that LOG=Y should be used for CLs which are notable/ should be on a changelog. I think some infra scripts are using this information but machenbach@ surely knows more about this.

[Jakob Kummerow]

Yeah, the ChangeLog is still generated from LOG=y commits: https://chromium.googlesource.com/v8/v8/+/branch-heads/4.8/ChangeLog 

(Note that the master branch's ChangeLog is not updated automatically; syncing it to the latest branch-head is a manual process that apparently hasn't been done in half a year.)

We've discussed getting rid of the ChangeLog repeatedly over the years. Nobody on our team needs/uses it, however there do seem to be some external people reading it; or at least that was the conclusion last time this came up. We could ask on v8-users if anyone would be unhappy if we dropped it, but that question doesn't necessarily reach everyone who might care, so it's hard to be sure.

--
--
v8-dev mailing list
v8-...@googlegroups.com
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to v8-dev+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Adam Klein]

Looking through that ChangeLog it doesn't look particularly useful (most releases have no log lines, though I suspect many of them do have significant changes).

What if we relaxed the presubmit to allow BUG= lines without LOG= lines? That would solve my main problem, while still letting people add LOG=Y lines if they think it's appropriate. The fact that the presubmit only complains for patches that have a BUG line seems like a perverse incentive to avoid using tracking bugs.

[Jakob Kummerow]

On Mon, Jan 11, 2016 at 8:29 PM, Adam Klein <ad...@chromium.org> wrote:

Looking through that ChangeLog it doesn't look particularly useful

Agreed.

 

(most releases have no log lines, though I suspect many of them do have significant changes).

I'm not so sure -- our tagged versions are really fine-grained these days, sometimes every single commit is tagged, so it could well be that many of those aren't ChangeLog-worthy (which of course doesn't help with the ChangeLog's lack of usefulness.)

What if we relaxed the presubmit to allow BUG= lines without LOG= lines? That would solve my main problem, while still letting people add LOG=Y lines if they think it's appropriate. The fact that the presubmit only complains for patches that have a BUG line seems like a perverse incentive to avoid using tracking bugs.

The rationale was that visible features should be tracked by bugs anyway, and requiring the LOG line forces people to make a decision whether a given change should be mentioned in the log or not, as we felt that we couldn't come up with a reasonable way to automate it.

I suspect that making the LOG line optional is pretty much equivalent to getting rid of it, as people will start to forget about it.

Which is fine with me :-)

[Daniel Ehrenberg]

On Mon, Jan 11, 2016 at 1:06 PM, Jakob Kummerow <jkum...@chromium.org> wrote:
> On Mon, Jan 11, 2016 at 8:29 PM, Adam Klein <ad...@chromium.org> wrote:
>>
>> Looking through that ChangeLog it doesn't look particularly useful
>
>
> Agreed.
>
>>
>> (most releases have no log lines, though I suspect many of them do have
>> significant changes).
>
>
> I'm not so sure -- our tagged versions are really fine-grained these days,
> sometimes every single commit is tagged, so it could well be that many of
> those aren't ChangeLog-worthy (which of course doesn't help with the
> ChangeLog's lack of usefulness.)
>
>> What if we relaxed the presubmit to allow BUG= lines without LOG= lines?
>> That would solve my main problem, while still letting people add LOG=Y lines
>> if they think it's appropriate. The fact that the presubmit only complains
>> for patches that have a BUG line seems like a perverse incentive to avoid
>> using tracking bugs.
>
>
> The rationale was that visible features should be tracked by bugs anyway,
> and requiring the LOG line forces people to make a decision whether a given
> change should be mentioned in the log or not, as we felt that we couldn't
> come up with a reasonable way to automate it.

What's an example of something that shouldn't be tracked by bugs? I
like to track everything that I'm doing by bugs, and find it easier to
understand others' work when it's organized that way too.

Dan

[Adam Klein]

Some stats on the ChangeLog in the 4.9 branch:

- Out of 385 minor revisions, 60 contained non-trivial ChangeLogs

- A total of 75 commits were mentioned in those 60 revisions

Does anyone here object to getting rid of the LOG= requirement? With fewer than 1 in 5 minor revisions containing a commit log, I highly doubt that it's a useful signal. I'll send a message out to v8-users unless I hear objections.

On Mon, Jan 11, 2016 at 1:06 PM, Jakob Kummerow <jkum...@chromium.org> wrote:

[Michael Hablich]

I agree with Dan. I think it make more sense to simply use a tracking bug. 

  +Michael Achenbach any objections?

[Adam Klein]

I'd like to separate the questions of requiring LOG= lines and requiring BUG= lines. I think the former is clearly not doing much for us (given my exploration of the data). For the latter, I think there's a variety of workflows used by folks on the team, and forcing tracking bugs for every patch seems like a major shift in process.

[Daniel Ehrenberg]

To clarify, I wasn't arguing that all of those workflows should have BUG= lines.

Dan

[Toon Verwaest]

Ping. Was just bitten by this again :)

[Adam Klein]

Sent out a proposed CL to remove the check to https://codereview.chromium.org/1956993002/

[Adam Klein]

The check has now been removed. Carry on adding BUG lines without getting nagged. Note that the ChangeLog still exists, so adding LOG=y will still put your change in the ChangeLog. Whether we should stop maintaining that altogether is a question for another day.

[Georg Neis]

Thanks Adam!