Re: [PATCH] Revert "x86/tsc: Consolidate init code"

[Ville Syrjälä]

On Mon, Sep 10, 2018 at 05:25:38PM +0200, Borislav Petkov wrote:
> On Mon, Sep 10, 2018 at 06:09:10PM +0300, Ville Syrjälä wrote:
> > But it is a patch, and if it happens to get accepted as is so be
> > it. If not, it's a good place where to start the conversation on
> > how to fix the bug in another way.
> 
> Uh, more of that "logic".
> 
> It is a patch but not really, if it is applied, good, if not, also good.
> WTF dude?
> 
> > You guys seem to have a notion that anything which says '[PATCH]'
> > is somehow final. In my book any patch is up for debate. Nothing
> > special about this one in that regard.
> 
> Well, let's see: imagine you're a maintainer. You get gazillion patches
> a day. And you think, oh well, I need to review and possibly apply this.
> And then move on to the next one. Because everyone is asking, when is
> she/he going to apply my damn patches...

Sounds to me like the maintainer should figure out how to delegate some
of the load a bit. Or just go on vacation and ignore all mails. I hear
stress isn't good for you.

> 
> But nooo, *some* of the patches are special - they're a conversation
> starter *only*! But also if applied, that's fine too.

That's what all patches are. No should be applying unreviewed patches
blindly.

Also often a revert is a perfect way to handle regressions. It gets
the angry users off your back ASAP allowing you to fix the bug
properly without having to rush it. I only wish all regression I've
caused would have been caught early enough for a revert to apply
cleanly. 

Even if the revert isn't applied the fact that the mail has the 
offending code right there makes the disussion easier. No need to
git fetch; git show <copy paste sha>; copy paste some code snippets
into the mail, etc.).

> 
> What a bunch of bull!

Calm down. No one is out to revert all your patches.

> 
> What's wrong with sending a mail tagged with "[REGRESSION]" - this looks
> like the tag people have adopted - and explain what the problem is, what
> you've bisected it to and what your observations are? Like everyone else
> reporting bugs/regressions/...

Maybe you can propose a new git-regression tool then? And document that
you want bugs reported using it? Ideally I'd say it should do almost
exactly what git revert does except s/revert/regression/. Though I
suppose it could include the original diff instead of the reverse.

Now, how about we stop this pointless "logic" discussion and
focus on the techinal stuff from now on?

-- 
Ville Syrjälä
Intel