On Mon, Oct 21, 2013 at 04:20:02PM -0700, Randy Dunlap wrote:
> On 10/21/13 08:36, Thierry Reding wrote:
> > Hi all,
> > 
> > I've uploaded today's linux-next tree to the master branch of the
> > repository below:
> > 
> >         git://gitorious.org/thierryreding/linux-next.git
> > 
> > A next-20131021 tag is also provided for convenience.
> > 
> > Gained a couple new conflicts, loads of build failures. I've tried to
> > fix the most obvious ones, but as a downside haven't gotten around to
> > write up the specifics about the conflicts. I might still do that
> > tomorrow, but I figured most people wouldn't have time to look into
> > those anyway given that they'll be busy with the conferences.
> 
> on i386:
> 
> drivers/built-in.o: In function `panel_simple_probe':
> panel-simple.c:(.text+0x3c7ca): undefined reference to `of_find_backlight_by_node'

Hi Randy,

Thanks for running this. I'll come up with a patch to fix this, since I
originally wrote of_find_backlight_by_node().

Running the randconfig that you provided yielded some other interesting
warnings, though. There are a few occurrences where gcc compiles about
functions not returning a value, which is caused by people assuming that
BUG() never returns. But it happens that your randconfig also unsets
CONFIG_BUG, which turns BUG into a no-op and therefore causing gcc to
rightfully complain.

Before attempting to fix these, I was wondering whether that would be
considered useful. I personally think that we should eliminate warnings
as much as possible because it helps filtering out genuine problems. But
the BUG Kconfig option also depends on EXPERT, so I guess it's usually
not even visible and possibly not worth bothering about.

Thierry