Build regressions/improvements in v4.3-rc3

Build regressions/improvements in v4.3-rc3

[Geert Uytterhoeven]

On Sun, Sep 27, 2015 at 9:34 PM, Geert Uytterhoeven
<geert@linux-m68k.org> wrote:
> JFYI, when comparing v4.3-rc3[1] to v4.3-rc2[3], the summaries are:
>   - build errors: +8/-12

  + error: No rule to make target include/config/auto.conf:  => N/A

arm-randconfig

> [1] http://kisskb.ellerman.id.au/kisskb/head/9404/ (255 out of 254 configs)
> [3] http://kisskb.ellerman.id.au/kisskb/head/9376/ (all 254 configs)

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

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

Build regressions/improvements in v4.3-rc3

[Russell King - ARM Linux]

On Mon, Sep 28, 2015 at 08:57:44AM +0200, Geert Uytterhoeven wrote:
> On Sun, Sep 27, 2015 at 9:34 PM, Geert Uytterhoeven
> <geert@linux-m68k.org> wrote:
> > JFYI, when comparing v4.3-rc3[1] to v4.3-rc2[3], the summaries are:
> >   - build errors: +8/-12
> 
>   + error: No rule to make target include/config/auto.conf:  => N/A
> 
> arm-randconfig

Not every randconfig failure is a kernel problem.  Here's an example:

Physical address of main memory (PHYS_OFFSET) [] (NEW) aborted!

Console input/output is redirected. Run 'make oldconfig' to update configuration.

This needs someone to provide a value, which means these failures are
not in fact failures of the kernel, but a failure of the build system
to anticipate that there may be Kconfig questions that need input.

So, these ones should be ignored.

-- 
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

Build regressions/improvements in v4.3-rc3

[Russell King - ARM Linux]

On Mon, Sep 28, 2015 at 10:30:06AM +0100, Russell King - ARM Linux wrote:
> On Mon, Sep 28, 2015 at 08:57:44AM +0200, Geert Uytterhoeven wrote:
> > On Sun, Sep 27, 2015 at 9:34 PM, Geert Uytterhoeven
> > <geert@linux-m68k.org> wrote:
> > > JFYI, when comparing v4.3-rc3[1] to v4.3-rc2[3], the summaries are:
> > >   - build errors: +8/-12
> > 
> >   + error: No rule to make target include/config/auto.conf:  => N/A
> > 
> > arm-randconfig
> 
> Not every randconfig failure is a kernel problem.  Here's an example:
> 
> Physical address of main memory (PHYS_OFFSET) [] (NEW) aborted!
> 
> Console input/output is redirected. Run 'make oldconfig' to update configuration.
> 
> This needs someone to provide a value, which means these failures are
> not in fact failures of the kernel, but a failure of the build system
> to anticipate that there may be Kconfig questions that need input.
> 
> So, these ones should be ignored.

Also, a great many of the failures are due to the build toolchain not
supporting -fstack-protector-strong.  I think it's absolutely right
for the build to error out if you enable a kernel feature which requires
toolchain support, but the toolchain does not support that feature.

What this means is that the build results from kissb are less than
useful - without spending ages looking at every single build, it's
hard to find the real failures we care about.

I'd suggest that either the randconfig is seeded to ensure that
CONFIG_CC_STACKPROTECTOR_STRONG is always disabled in ARM randconfigs,
or that the ARM toolchain is updated to support this feature.

-- 
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

Build regressions/improvements in v4.3-rc3

[Geert Uytterhoeven]

Hi Russell,

On Mon, Sep 28, 2015 at 11:48 AM, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
> On Mon, Sep 28, 2015 at 10:30:06AM +0100, Russell King - ARM Linux wrote:
>> On Mon, Sep 28, 2015 at 08:57:44AM +0200, Geert Uytterhoeven wrote:
>> > On Sun, Sep 27, 2015 at 9:34 PM, Geert Uytterhoeven
>> > <geert@linux-m68k.org> wrote:
>> > > JFYI, when comparing v4.3-rc3[1] to v4.3-rc2[3], the summaries are:
>> > >   - build errors: +8/-12
>> >
>> >   + error: No rule to make target include/config/auto.conf:  => N/A
>> >
>> > arm-randconfig
>>
>> Not every randconfig failure is a kernel problem.  Here's an example:
>>
>> Physical address of main memory (PHYS_OFFSET) [] (NEW) aborted!
>>
>> Console input/output is redirected. Run 'make oldconfig' to update configuration.
>>
>> This needs someone to provide a value, which means these failures are
>> not in fact failures of the kernel, but a failure of the build system
>> to anticipate that there may be Kconfig questions that need input.
>>
>> So, these ones should be ignored.

Usually I ignore them. But as this was the only failure to report (I ignore
R_PPC64_REL24 failures, too), I thought to kick the kbuild people (and the
ARM list, as it was the ARM randconfig that failed) for once.

> Also, a great many of the failures are due to the build toolchain not
> supporting -fstack-protector-strong.  I think it's absolutely right
> for the build to error out if you enable a kernel feature which requires
> toolchain support, but the toolchain does not support that feature.
>
> What this means is that the build results from kissb are less than
> useful - without spending ages looking at every single build, it's
> hard to find the real failures we care about.
>
> I'd suggest that either the randconfig is seeded to ensure that
> CONFIG_CC_STACKPROTECTOR_STRONG is always disabled in ARM randconfigs,
> or that the ARM toolchain is updated to support this feature.

Or that it's impossible to enable this feature if your toolchain doesn't
support it?

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

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

Build regressions/improvements in v4.3-rc3

[Russell King - ARM Linux]

On Mon, Sep 28, 2015 at 03:50:16PM +0200, Geert Uytterhoeven wrote:
> Hi Russell,
> 
> On Mon, Sep 28, 2015 at 11:48 AM, Russell King - ARM Linux
> <linux@arm.linux.org.uk> wrote:
> > Also, a great many of the failures are due to the build toolchain not
> > supporting -fstack-protector-strong.  I think it's absolutely right
> > for the build to error out if you enable a kernel feature which requires
> > toolchain support, but the toolchain does not support that feature.
> >
> > What this means is that the build results from kissb are less than
> > useful - without spending ages looking at every single build, it's
> > hard to find the real failures we care about.
> >
> > I'd suggest that either the randconfig is seeded to ensure that
> > CONFIG_CC_STACKPROTECTOR_STRONG is always disabled in ARM randconfigs,
> > or that the ARM toolchain is updated to support this feature.
> 
> Or that it's impossible to enable this feature if your toolchain doesn't
> support it?

I think that would require kconfig to call out to the compiler to run
a test build.  This kind of thing has been talked about in the past
but the feature never seems to appear - probably because it would end
up increasing the kconfig startup time in every case.

-- 
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

Build regressions/improvements in v4.3-rc3

[Geert Uytterhoeven]

Hi Russell,

On Mon, Sep 28, 2015 at 3:58 PM, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
> On Mon, Sep 28, 2015 at 03:50:16PM +0200, Geert Uytterhoeven wrote:
>> On Mon, Sep 28, 2015 at 11:48 AM, Russell King - ARM Linux
>> <linux@arm.linux.org.uk> wrote:
>> > Also, a great many of the failures are due to the build toolchain not
>> > supporting -fstack-protector-strong.  I think it's absolutely right
>> > for the build to error out if you enable a kernel feature which requires
>> > toolchain support, but the toolchain does not support that feature.
>> >
>> > What this means is that the build results from kissb are less than
>> > useful - without spending ages looking at every single build, it's
>> > hard to find the real failures we care about.
>> >
>> > I'd suggest that either the randconfig is seeded to ensure that
>> > CONFIG_CC_STACKPROTECTOR_STRONG is always disabled in ARM randconfigs,
>> > or that the ARM toolchain is updated to support this feature.
>>
>> Or that it's impossible to enable this feature if your toolchain doesn't
>> support it?
>
> I think that would require kconfig to call out to the compiler to run
> a test build.  This kind of thing has been talked about in the past
> but the feature never seems to appear - probably because it would end
> up increasing the kconfig startup time in every case.

Given the number of existing "$(call cc-option, ...)" in Makefiles, the effect
on kconfig startup time should be small.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

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