On Thursday, May 01, 2014 12:38:28 PM Hans de Goede wrote:

Hi,

On 04/30/2014 09:52 PM, Rafael J. Wysocki wrote:

quoted

On Wednesday, April 30, 2014 03:37:21 PM Hans de Goede wrote:

quoted

This fixes the backlight control not working.

Cc: stable@vger.kernel.org
Reported-and-tested-by: Vincent Gerris <redacted>
Signed-off-by: Hans de Goede <redacted>

Sorry, this conflicts with commit 170269a9d3c0 (ACPI / video: Default to using
native backlight control on Windows 8 systems) in linux-next, so I'm not going
to apply it.

I strongly disagree, rejecting bug-fixes which conflict with more rigorous
(and dangerous) fixes -next, purely because the conflict with something -next
is not a good reason. TBH I find it a complete non reason to reject these fixes.

quoted

If you wanted to have this stuff in 3.15, there was a plenty of time to submit
it earlier.

Heh, that assumes I was aware of this particular model needing this quirk earlier,
while I actually got the first report of it not working from Vincent on April 26th,
and got confirmation that the quirk fixes it on April 29th. I would say that 1 day
turn around time between getting the confirmation and sending the patch is not bad
at all.

I really believe it is important to get the quirk for this model (and others) into
3.15, here us my decision tree leading to this:

-Do we want to fix these brightness issues -> Yes
-Do we expect our users to wait for 6 months for an upstream fix + many more months
 for the fixed kernel to hit distros -> No
-So we want to backport these fixes to stable -> Yes
-Is the proposed fix for 3.16 acceptable for stable -> No (too high change of
 regressions)

OK, this is a good argument.

Conclusion: we want quirks for models known to need quirks added to 3.15 and
backported to the various stable series.

I actually want to go as far as to claim that once 3.15 is released we will want
to add quirks to 3.15.x, breaking the every fix must be upstream rule for the stable
series. But lets safe that discussion for later.

Well, there's a way out of this.  Instead of doing commit 170269a9d3c0 as is, we
can just switch the default without removing the blacklist just yet.  And remove
the blacklist one we are reasonably confident that the new default actually works.

In which case I'd go for your original series (along with the RFC moving stuff
out of blacklist.c to the video.c blacklist) with a replacement of commit
170269a9d3c0 that will simply flip the default.

Does this make sense to you?

I've been doing triaging of (Fedora) brightness bugs for the last few days (I sort of
just rolled into this whole brightness business) and this really is a big problem,

I'm very much aware of that.

Thanks,
Rafael