Hi Günter,
On Wed, Feb 5, 2020 at 2:52 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
> On 2/5/20 1:03 AM, Geert Uytterhoeven wrote:
> > On Wed, Feb 5, 2020 at 4:57 AM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
> >> On 2/4/20 7:34 PM, Dan Carpenter wrote:
> >>> On Tue, Feb 04, 2020 at 12:31:16PM -0800, Matthew Wilcox wrote:
> >>>> On Tue, Feb 04, 2020 at 08:06:14PM +0000, Chris Packham wrote:
> >>>>> On Tue, 2020-02-04 at 07:09 +0000, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> >>>>>> On Tue, Feb 04, 2020 at 04:02:15AM +0000, Chris Packham wrote:
> >>>>> On Tue, 2020-02-04 at 10:21 +0300, Dan Carpenter wrote:
> >>>>>> My advice is to delete all the COMPILE_TEST code. That stuff was a
> >>>>>> constant source of confusion and headaches.
> >>>>>
> >>>>> I was also going to suggest this. Since the COMPILE_TEST has been a
> >>>>> source of trouble I was going to propose dropping the || COMPILE_TEST
> >>>>> from the Kconfig for the octeon drivers.
> >>>>
> >>>> Not having it also causes problems. I didn't originally add it for
> >>>> shits and giggles.
> >>>
> >>> I wonder if the kbuild bot does enough cross compile build testing these
> >>> days to detect compile problems. It might have improved to the point
> >>> where COMPILE_TEST isn't required.
> >
> > It depends...
> >
> >> Not really. Looking at the build failures in the mainline kernel right now:
> >>
> >> Failed builds:
> >> alpha:allmodconfig
> >> arm:allmodconfig
> >> i386:allyesconfig
> >> i386:allmodconfig
> >> m68k:allmodconfig
> >> microblaze:mmu_defconfig
> >> mips:allmodconfig
> >> parisc:allmodconfig
> >> powerpc:allmodconfig
> >> s390:allmodconfig
> >> sparc64:allmodconfig
> >
> > I did receive a report from noreply@xxxxxxxxxxxxxx for the m68k build
> > failure. But that was sent to me only, not to the offender, and I do my
> > own builds anyway.
> >
> > More interesting, that report happened after the offending commit landed
> > upstream, while it had been in next for 4 weeks.
>
> m68k in -next builds fine for me, and did for a while. I have not seen a build
> failure there. There must be a context commit causing this failure, or what
> is (or was) in -next differs from what is in mainline.
Indeed. The offending symbol depended on another symbol, which never
made it to next before it hit mainline, thus hiding the issue.
> >> Many of those don't even _have_ specific configurations causing the build failures.
> >
> > Exactly. These are the "easy" ones, as the all*config builds enable as
> > much infrastructure as possible. It's much harder if some common
> > dependency is not fulfilled in some specific config.
>
> Yes, that is correct. But that doesn't mean that it would be a good idea
> to retire COMPILE_TEST.
I agree.
Retiring COMPILE_TEST could have a positive side-effect, though: it
would reduce the compile time for all*config, which might give the
overloaded build bots spare cycles to cover other configs.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
