Mailing List Archive: "Trivial" commit reviews

"Trivial" commit reviews

Sep 23, 2007, 3:40 PM

Post #1 of 7 (170 views)

Mike Doty (KingTaco) just told me I could stop sending reviews to -dev
that are just about adding quotes or other trivial issues that come up
over and over. I'm going to tell you why it's still a good thing.

First, where one problem lurks, others often do too. In code with such
simple problems, it's likely that more complex problems also exist.
Getting more eyes on problematic code of any sort can help find them.

Second, as we've already seen, no one developer is familiar with all the
code. Both Mike Frysinger and Daniel Drake have responded to some of my
reviews, pointing out further problems with the same code.

Third, by continuing to post these reviews, it should become obvious to
_all_ developers that they should be checking for them _before_
committing instead of waiting for a review.

Over time, the number of these simple reviews should go dramatically
down so it no longer bothers anyone to see them. If it doesn't, that
means some of our developers aren't learning or paying attention, and we
should take a closer look at whether they should remain developers.

Thanks,
Donnie
--
gentoo-dev@gentoo.org mailing list

Re: "Trivial" commit reviews [ In reply to ]

kingtaco at gentoo

Sep 23, 2007, 4:02 PM

Post #2 of 7 (161 views)

Permalink

Donnie Berkholz wrote:
> Mike Doty (KingTaco) just told me I could stop sending reviews to -dev
> that are just about adding quotes or other trivial issues that come up
> over and over. I'm going to tell you why it's still a good thing.
>
> First, where one problem lurks, others often do too. In code with such
> simple problems, it's likely that more complex problems also exist.
> Getting more eyes on problematic code of any sort can help find them.
>
> Second, as we've already seen, no one developer is familiar with all the
> code. Both Mike Frysinger and Daniel Drake have responded to some of my
> reviews, pointing out further problems with the same code.
>
> Third, by continuing to post these reviews, it should become obvious to
> _all_ developers that they should be checking for them _before_
> committing instead of waiting for a review.
>
> Over time, the number of these simple reviews should go dramatically
> down so it no longer bothers anyone to see them. If it doesn't, that
> means some of our developers aren't learning or paying attention, and we
> should take a closer look at whether they should remain developers.
>
> Thanks,
> Donnie
My concern is that if we flood -dev with "trivial" commit problems then
more people will stop watching -dev and/or resort to killfiles or other
filtering. While I do agree with Donnies assessment, my concern is that
over a longer time period, it might have a negative effect.

--
=======================================================
Mike Doty kingtaco -at- gentoo.org
Gentoo Infrastructure
Gentoo/AMD64 Strategic Lead
GPG: E1A5 1C9C 93FE F430 C1D6 F2AF 806B A2E4 19F4 AE05
=======================================================
--
gentoo-dev@gentoo.org mailing list

Re: "Trivial" commit reviews [ In reply to ]

bangert at gentoo

Sep 23, 2007, 9:22 PM

Post #3 of 7 (166 views)

Permalink

i am all for the 'trivial' review. as i am not on the commit list,
however, i can't tell whether this acutally helps.

do people fix the stuff that is pointed out to them?

also, perhaps the more common ones should additionally be converted to
repoman tests, if that is feasable.

kind regards
Thilo

Re: "Trivial" commit reviews [ In reply to ]

dberkholz at gentoo

Sep 23, 2007, 9:44 PM

Post #4 of 7 (164 views)

Permalink

On 06:22 Mon 24 Sep , Thilo Bangert wrote:
> do people fix the stuff that is pointed out to them?

Yep, I've seen a lot of fixes for reviews.

> also, perhaps the more common ones should additionally be converted to
> repoman tests, if that is feasable.

That might be reasonable for some cases, but it won't be perfect, and
won't even be possible for many.

So far, the only one I've seen that might work well for is quoting
around specific variables. You could do something like a grep for words
containing '${+D[^[:alnum:]-_]' (haven't tested that, just beginnings of
an idea) and the same for S and WORKDIR.

Thanks,
Donnie
--
gentoo-dev@gentoo.org mailing list

Re: "Trivial" commit reviews [ In reply to ]

mabi at gentoo

Sep 24, 2007, 12:26 AM

Post #5 of 7 (166 views)

Permalink

Mike Doty <kingtaco@gentoo.org> wrote:
> Donnie Berkholz wrote:
>> Over time, the number of these simple reviews should go dramatically down
>> so it no longer bothers anyone to see them. If it doesn't, that means some
>> of our developers aren't learning or paying attention, and we should take
>> a closer look at whether they should remain developers.

> My concern is that if we flood -dev with "trivial" commit problems then
> more people will stop watching -dev and/or resort to killfiles or other
> filtering. While I do agree with Donnies assessment, my concern is that
> over a longer time period, it might have a negative effect.

I totally agree with Donnie here. Please keep up the work, everybody
should be encouraged to fix these (trivial) problems. I sincerly hope
that these message will not have to continue for long. But as long as
they do, they serve as a big reminder in your inbox of what is wrong.

Just my 0.02$
--
Regards, Matti Bickel
Signed/Encrypted email preferred (key 4849EC6C)

Re: "Trivial" commit reviews [ In reply to ]

remi at gentoo

Sep 24, 2007, 3:15 PM

Post #6 of 7 (167 views)

Permalink

Matti Bickel wrote:
> I totally agree with Donnie here. Please keep up the work, everybody
> should be encouraged to fix these (trivial) problems. I sincerly hope
> that these message will not have to continue for long. But as long as
> they do, they serve as a big reminder in your inbox of what is wrong.

+1, I've already learned or re-learned some very useful stuff.

/me will keep reading those ebuild reviews.

Rémi
--
gentoo-dev@gentoo.org mailing list

Re: "Trivial" commit reviews [ In reply to ]

dirtyepic at gentoo

Jun 17, 2008, 1:23 AM

Post #7 of 7 (163 views)

Permalink

Mike Doty wrote:
> Donnie Berkholz wrote:
>> Mike Doty (KingTaco) just told me I could stop sending reviews to -dev
>> that are just about adding quotes or other trivial issues that come up
>> over and over. I'm going to tell you why it's still a good thing.
>>
>> First, where one problem lurks, others often do too. In code with such
>> simple problems, it's likely that more complex problems also exist.
>> Getting more eyes on problematic code of any sort can help find them.
>>
>> Second, as we've already seen, no one developer is familiar with all
>> the code. Both Mike Frysinger and Daniel Drake have responded to some
>> of my reviews, pointing out further problems with the same code.
>>
>> Third, by continuing to post these reviews, it should become obvious
>> to _all_ developers that they should be checking for them _before_
>> committing instead of waiting for a review.
>>
>> Over time, the number of these simple reviews should go dramatically
>> down so it no longer bothers anyone to see them. If it doesn't, that
>> means some of our developers aren't learning or paying attention, and
>> we should take a closer look at whether they should remain developers.
>>
>> Thanks,
>> Donnie
> My concern is that if we flood -dev with "trivial" commit problems then
> more people will stop watching -dev and/or resort to killfiles or other
> filtering. While I do agree with Donnies assessment, my concern is that
> over a longer time period, it might have a negative effect.

I think this is exactly what -dev is for. And the number of trivial
mistakes people are making (not discounting myself) shows this kind of code
review is needed. There is a huge amount of basic stuff, like quoting or
when to use ${ROOT}, that we just don't have documented anywhere, and expect
people to somehow just pick up.

If we flood -dev with "trivial" commit problems then more people will stop
commiting them. If people want to stop reading or filter then that's their
prerogative. -dev isn't a required ML.

--
fonts / wxWindows / gcc-porting / treecleaners
9B81 6C9F E791 83BB 3AB3 5B2D E625 A073 8379 37E8 (0x837937E8)

--
gentoo-dev@gentoo.org mailing list

Mailing List Archive

Mailing List Archive

Attached Files: