Message-ID: <20150624131502.GE32642@pd.tnic>
Date:	Wed, 24 Jun 2015 15:15:03 +0200
From:	Borislav Petkov <bp@...e.de>
To:	Linus Torvalds <torvalds@...ux-foundation.org>
Cc:	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	linux-edac <linux-edac@...r.kernel.org>,
	the arch/x86 maintainers <x86@...nel.org>
Subject: Re: [GIT PULL] EDAC updates for 4.

On Wed, Jun 24, 2015 at 06:01:41AM -0700, Linus Torvalds wrote:
> On Wed, Jun 24, 2015 at 5:40 AM, Linus Torvalds
> <torvalds@...ux-foundation.org> wrote:
> >
> > You didn't actually test what you sent me. YOU TESTED SOMETHING
> > ENTIRELY DIFFERENT.
> 
> Btw, it worries me that not only are you in denial about this,
> apparently you have always done it:
> 
>   "But I have always merged the tip/x86/ras branch which contained the x86
>    changes into the EDAC tree when testing. Basically what I should've done
>    with the pull request too"

"always" meant during the 4.1-rc cycle, of course. Only for this
release.

> because this shows that your workflow is just fundamentally broken.
> 
> You should test *YOUR* branch. That's the primary thing. Make sure
> your code works and makes sense, and nothing else really matters.
>
> Sure, feel free to go ahead and also test whatever other combinations
> (more testing is never wrong), but those are very definitely
> secondary, and aren't nearly as important. And it is never a
> _replacement_ for testing your branch, it is always a "in addition
> to".

Ok, understood.

> I'd much rather you test the thing you send me twice as much, and
> *never* test any combination, than seeing that you primarily test
> combinations with other branches.
> 
> And yes, if it then turns out that there are often interactions with
> other branches that means that the integrated thing doesn't work (even
> after the individual branches have been tested extensively and work on
> their own), then sure, that can be a problem.
> 
> Those kinds of problems are fairly unusual, but they tend to mean that
> multiple people are stepping on each others toes. It isn't all that
> different from "those two development trees often cause conflicts",
> and usually means that either the code needs some re-organization so
> that people can work better independently, or it means that the
> different branches really are working on the same thing, and perhaps
> need to be working more closely together.
> 
> But generally, the *less* intertwined you are, the better off you are.
> It's usually much better to try to have different branches and
> developers be as independent as possible, so that they don't get
> serialization issues.

Yeah, so as I said earlier, in hindsight, I should've stuck the error
injection stuff completely into tip as it depends on it. But we carry it
in drivers/edac/ for some archaic reason or because it was easier this
way at the time. In the meantime, it depends so much on x86 facilities
that it actually belongs into arch/x86/ras/. I even had patches which
did that.

I'll try to dust them off for 4.2-rc maybe, let's see what happens.

In the meantime, I've zapped those two offending patches and am testing
a v2 pull request.

Thanks for explaining the situation, fully agreed and noted for the
future.

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists