[Mesa-dev] [PATCH v2 1/2] mesa: Add KBL PCI IDs and platform information.

[Ben Widawsky]

On Fri, Nov 20, 2015 at 02:18:56PM +0000, Emil Velikov wrote:
> On 18 November 2015 at 21:59, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
> > On Wed, Nov 18, 2015 at 4:54 PM, Sarah Sharp
> > <sarah.a.sharp at linux.intel.com> wrote:
> >>> There's not really a consensus I guess, but most people do leave the version
> >>> information in the final commit message.
> >>
> >> I personally feel like that's leaving boredom doodles on a final
> >> architectural drawing. If people want to know the back-and-forth
> >> history, the mailing list archive will always be there. So, no, I don't
> >> really want to leave version info in the commit message.
> >
> > FWIW I wholeheartedly agree with this line of reasoning. I never put
> > the version info into my commits either, and find it
> > confusing/misleading when others do. I want to know the final state of
> > things when looking at the commit 1 year from now, not the 20-step
> > process and all the wrong turns to get there.

When I'm lazy, the commit message doesn't reflect the final state and you need
to read through the version info to get the end result. That is not great
because I am usually lazy. However aside from that case, I don't see what is
confusing about it - it's at the bottom. Don't read it if you don't want to.

> >
> 
> The other side of the coin:
> 
>  - One might not have access to the discussion - ISP/ML archive is
> down. discussion was offline or no longer available (10+ years ago),
> etc.
>  - Revision history is immediately available, rather than going back
> and forth between git/browser/email client.
>  - We can easily ignore the revision history hunk
>  - Hitting more than v3 is a clear sign something fishy (most likely
> lack of experience of said author), which in itself is useful.

I agree with all but this one - and of course I am partial because I almost
never land something in under v3. It could also mean there was a lot of
contention about something, or a poor review process (for instance if v5-v10 are
all comment changes or commit message changes, that is not an efficient review).
I think you do hit on a reason some people might not want to put version
information in - because there is some amount of shaming and embarrassment.

> 
> Obviously I'm not saying that we must use approach A or B. Whatever
> floats your boat really :-)
> 
> -Emil

I was forced to do the version thing against my will originally for the kernel
driver, but over time it grew on me. (I wanted to do it as git notes, but that
was rejected).

I personally like the version information because a lot of times the changes
that go in there are things which I don't necessarily want or like, but was
requested of me to make in order to get a reviewed-by. As a result, when I look
back and say, WTF did I do that - it's immediately obvious without having to
look at the m-l. Similarly, I tend to do the most amount of testing on v1.
There have been several times where my patch regressed something, and it's
useful to try to figure out how I could have missed such a regression - the
version information often has the answer.

What is definitely clear is there is not consensus, and I even take back my
original assertion that "most" people do this. Going through the history I found
that there are some developers who themselves don't have a reliable approach.