[U-Boot] Treat warnings different from errors in buildman?

[U-Boot] Treat warnings different from errors in buildman?

[Tom Rini]

Hey,

So as I migrate scripts over to buildman, one issue I have is that
today warning and errors are treated the same:
u-boot (master)$ ./tools/buildman/buildman -b master -c 1 -ve -T 1 -j 9 'arc|blackfin|microblaze|m68k|nds32|sparc|x86|aarch64|sandbox|mips' -s
Summary of 1 commit for 128 boards (1 thread, 9 jobs per thread)
01: git-mailrc: add patman and buildman alias
      mips: +   pb1000 dbau1550_el qemu_mipsel maltael
     sparc: +   grsim_leon2 gr_cpci_ax2000 gr_ep2s60 grsim gr_xc3s_1500
       arc: +   tb100 arcangel4-be axs101 arcangel4
   sandbox: +   sandbox

mips is the mipsel libgcc thing, I need to see if eldk5.5 behaves
better than 5.2.1 in that regard and passing or setting
CONFIG_USE_PRIVATE_LIBGCC=yes.  sparc is however all warnings (old gcc
warning when it should not).

Also, is there any way to note which boards/arches an error comes from?
+/home/trini/work/u-boot/master/.bm-work/00/disk/part.c: In function
`get_device':
+/home/trini/work/u-boot/master/.bm-work/00/disk/part.c:453: warning:
'hwpart' might be used uninitialized in this function

Is confusing if you hadn't done a sparc solo build before and seen that.

Thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20140821/b6c49953/attachment.pgp>

[U-Boot] Treat warnings different from errors in buildman?

[Tom Rini]

On Thu, Aug 21, 2014 at 12:53:42PM -0400, Tom Rini wrote:

> Hey,
> 
> So as I migrate scripts over to buildman, one issue I have is that
> today warning and errors are treated the same:
> u-boot (master)$ ./tools/buildman/buildman -b master -c 1 -ve -T 1 -j 9 'arc|blackfin|microblaze|m68k|nds32|sparc|x86|aarch64|sandbox|mips' -s
> Summary of 1 commit for 128 boards (1 thread, 9 jobs per thread)
> 01: git-mailrc: add patman and buildman alias
>       mips: +   pb1000 dbau1550_el qemu_mipsel maltael
>      sparc: +   grsim_leon2 gr_cpci_ax2000 gr_ep2s60 grsim gr_xc3s_1500
>        arc: +   tb100 arcangel4-be axs101 arcangel4
>    sandbox: +   sandbox

Oh, and the sandbox error is because with sparse installed, it picks
'cgcc' as the toolchain to use rather than 'gcc' and then fails to link
as there's no 'cld'.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20140821/c43592ba/attachment.pgp>

[U-Boot] Treat warnings different from errors in buildman?

[Daniel Schwierzeck]

Hi Tom,

On 21.08.2014 18:53, Tom Rini wrote:
> Hey,
> 
> So as I migrate scripts over to buildman, one issue I have is that
> today warning and errors are treated the same:
> u-boot (master)$ ./tools/buildman/buildman -b master -c 1 -ve -T 1 -j 9 'arc|blackfin|microblaze|m68k|nds32|sparc|x86|aarch64|sandbox|mips' -s
> Summary of 1 commit for 128 boards (1 thread, 9 jobs per thread)
> 01: git-mailrc: add patman and buildman alias
>       mips: +   pb1000 dbau1550_el qemu_mipsel maltael
>      sparc: +   grsim_leon2 gr_cpci_ax2000 gr_ep2s60 grsim gr_xc3s_1500
>        arc: +   tb100 arcangel4-be axs101 arcangel4
>    sandbox: +   sandbox
> 
> mips is the mipsel libgcc thing, I need to see if eldk5.5 behaves
> better than 5.2.1 in that regard and passing or setting
> CONFIG_USE_PRIVATE_LIBGCC=yes.  

ELDK-5.5 doesn't behave better. You will always need
CONFIG_USE_PRIVATE_LIBGCC=yes to build all MIPS boards at once and if
you don't have a full multilib toolchain. I'm still wondering why
Wolfgang decided long time ago, that U-Boot should rely on the
toolchain's libgcc and not use its own variant like the kernel does.

-- 
- Daniel

[U-Boot] Treat warnings different from errors in buildman?

[Tom Rini]

On Fri, Aug 22, 2014 at 11:01:52AM +0200, Daniel Schwierzeck wrote:
> Hi Tom,
> 
> On 21.08.2014 18:53, Tom Rini wrote:
> > Hey,
> > 
> > So as I migrate scripts over to buildman, one issue I have is that
> > today warning and errors are treated the same:
> > u-boot (master)$ ./tools/buildman/buildman -b master -c 1 -ve -T 1 -j 9 'arc|blackfin|microblaze|m68k|nds32|sparc|x86|aarch64|sandbox|mips' -s
> > Summary of 1 commit for 128 boards (1 thread, 9 jobs per thread)
> > 01: git-mailrc: add patman and buildman alias
> >       mips: +   pb1000 dbau1550_el qemu_mipsel maltael
> >      sparc: +   grsim_leon2 gr_cpci_ax2000 gr_ep2s60 grsim gr_xc3s_1500
> >        arc: +   tb100 arcangel4-be axs101 arcangel4
> >    sandbox: +   sandbox
> > 
> > mips is the mipsel libgcc thing, I need to see if eldk5.5 behaves
> > better than 5.2.1 in that regard and passing or setting
> > CONFIG_USE_PRIVATE_LIBGCC=yes.  
> 
> ELDK-5.5 doesn't behave better. You will always need
> CONFIG_USE_PRIVATE_LIBGCC=yes to build all MIPS boards at once and if
> you don't have a full multilib toolchain. I'm still wondering why
> Wolfgang decided long time ago, that U-Boot should rely on the
> toolchain's libgcc and not use its own variant like the kernel does.

Ah, what I mean was that I can't seem to find the right 'trick' to have
CONFIG_USE_PRIVATE_LIBGCC work today.  I tried make-flags for buildman
and export CONFIG_USE_PRIVATE_LIBGCC=yes in my wrapper around MAKEALL.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20140822/7a0b639d/attachment.pgp>

[U-Boot] Treat warnings different from errors in buildman?

[Daniel Schwierzeck]

On 22.08.2014 12:47, Tom Rini wrote:
> On Fri, Aug 22, 2014 at 11:01:52AM +0200, Daniel Schwierzeck wrote:
>> Hi Tom,
>>
>> On 21.08.2014 18:53, Tom Rini wrote:
>>> Hey,
>>>
>>> So as I migrate scripts over to buildman, one issue I have is that
>>> today warning and errors are treated the same:
>>> u-boot (master)$ ./tools/buildman/buildman -b master -c 1 -ve -T 1 -j 9 'arc|blackfin|microblaze|m68k|nds32|sparc|x86|aarch64|sandbox|mips' -s
>>> Summary of 1 commit for 128 boards (1 thread, 9 jobs per thread)
>>> 01: git-mailrc: add patman and buildman alias
>>>       mips: +   pb1000 dbau1550_el qemu_mipsel maltael
>>>      sparc: +   grsim_leon2 gr_cpci_ax2000 gr_ep2s60 grsim gr_xc3s_1500
>>>        arc: +   tb100 arcangel4-be axs101 arcangel4
>>>    sandbox: +   sandbox
>>>
>>> mips is the mipsel libgcc thing, I need to see if eldk5.5 behaves
>>> better than 5.2.1 in that regard and passing or setting
>>> CONFIG_USE_PRIVATE_LIBGCC=yes.  
>>
>> ELDK-5.5 doesn't behave better. You will always need
>> CONFIG_USE_PRIVATE_LIBGCC=yes to build all MIPS boards at once and if
>> you don't have a full multilib toolchain. I'm still wondering why
>> Wolfgang decided long time ago, that U-Boot should rely on the
>> toolchain's libgcc and not use its own variant like the kernel does.
> 
> Ah, what I mean was that I can't seem to find the right 'trick' to have
> CONFIG_USE_PRIVATE_LIBGCC work today.  I tried make-flags for buildman
> and export CONFIG_USE_PRIVATE_LIBGCC=yes in my wrapper around MAKEALL.
> 

The trick is to use "CONFIG_USE_PRIVATE_LIBGCC=y" instead of
"CONFIG_USE_PRIVATE_LIBGCC=yes" because it is handled like a boolean
Kconfig symbol.

BTW: I'm using env scripts for different toolchains which could be
sourced. For ELDK-5.5 I have:

$ cat env-eldk5.5-mips
export CONFIG_USE_PRIVATE_LIBGCC=y
export USE_PRIVATE_LIBGCC=yes
export
CROSS_COMPILE=/opt/eldk-5.5/mips/sysroots/i686-eldk-linux/usr/bin/mips-linux/mips-linux-

-- 
- Daniel

[U-Boot] Treat warnings different from errors in buildman?

[Tom Rini]

On Fri, Aug 22, 2014 at 01:11:08PM +0200, Daniel Schwierzeck wrote:

> On 22.08.2014 12:47, Tom Rini wrote:
> > On Fri, Aug 22, 2014 at 11:01:52AM +0200, Daniel Schwierzeck wrote:
> >> Hi Tom,
> >>
> >> On 21.08.2014 18:53, Tom Rini wrote:
> >>> Hey,
> >>>
> >>> So as I migrate scripts over to buildman, one issue I have is that
> >>> today warning and errors are treated the same:
> >>> u-boot (master)$ ./tools/buildman/buildman -b master -c 1 -ve -T 1 -j 9 'arc|blackfin|microblaze|m68k|nds32|sparc|x86|aarch64|sandbox|mips' -s
> >>> Summary of 1 commit for 128 boards (1 thread, 9 jobs per thread)
> >>> 01: git-mailrc: add patman and buildman alias
> >>>       mips: +   pb1000 dbau1550_el qemu_mipsel maltael
> >>>      sparc: +   grsim_leon2 gr_cpci_ax2000 gr_ep2s60 grsim gr_xc3s_1500
> >>>        arc: +   tb100 arcangel4-be axs101 arcangel4
> >>>    sandbox: +   sandbox
> >>>
> >>> mips is the mipsel libgcc thing, I need to see if eldk5.5 behaves
> >>> better than 5.2.1 in that regard and passing or setting
> >>> CONFIG_USE_PRIVATE_LIBGCC=yes.  
> >>
> >> ELDK-5.5 doesn't behave better. You will always need
> >> CONFIG_USE_PRIVATE_LIBGCC=yes to build all MIPS boards at once and if
> >> you don't have a full multilib toolchain. I'm still wondering why
> >> Wolfgang decided long time ago, that U-Boot should rely on the
> >> toolchain's libgcc and not use its own variant like the kernel does.
> > 
> > Ah, what I mean was that I can't seem to find the right 'trick' to have
> > CONFIG_USE_PRIVATE_LIBGCC work today.  I tried make-flags for buildman
> > and export CONFIG_USE_PRIVATE_LIBGCC=yes in my wrapper around MAKEALL.
> > 
> 
> The trick is to use "CONFIG_USE_PRIVATE_LIBGCC=y" instead of
> "CONFIG_USE_PRIVATE_LIBGCC=yes" because it is handled like a boolean
> Kconfig symbol.
> 
> BTW: I'm using env scripts for different toolchains which could be
> sourced. For ELDK-5.5 I have:
> 
> $ cat env-eldk5.5-mips
> export CONFIG_USE_PRIVATE_LIBGCC=y
> export USE_PRIVATE_LIBGCC=yes
> export
> CROSS_COMPILE=/opt/eldk-5.5/mips/sysroots/i686-eldk-linux/usr/bin/mips-linux/mips-linux-

Ah, OK.  I've got something started on buildman now for an env-flags
section and it's building mips/mipsel.  I'll post a patch once it's
less embarassing looking.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20140822/87042d29/attachment.pgp>

[U-Boot] Treat warnings different from errors in buildman?

[Simon Glass]

Hi Tom,

On 21 August 2014 10:53, Tom Rini <trini@ti.com> wrote:
> Hey,
>
> So as I migrate scripts over to buildman, one issue I have is that
> today warning and errors are treated the same:

Actually buildman can't tell the difference at present. Both errors
and warnings end up in stderr. Buildman assumes that if make returns
an exit code of 0 then there were no errors (since they would have
terminated the build). So we have three categories:

- board with no errors/warnings
- board which builds but has output in stderr, assumed to be warnings
- board with errors, but may also contain warnings

I will take a look at trying to distinguish between warnings and
errors by looking at the stderr output. My philosophy with buildman
has been to go to quite extreme measured to help summarise and
categorise the problems, so I suppose this would be just another step.

> u-boot (master)$ ./tools/buildman/buildman -b master -c 1 -ve -T 1 -j 9 'arc|blackfin|microblaze|m68k|nds32|sparc|x86|aarch64|sandbox|mips' -s
> Summary of 1 commit for 128 boards (1 thread, 9 jobs per thread)
> 01: git-mailrc: add patman and buildman alias
>       mips: +   pb1000 dbau1550_el qemu_mipsel maltael
>      sparc: +   grsim_leon2 gr_cpci_ax2000 gr_ep2s60 grsim gr_xc3s_1500
>        arc: +   tb100 arcangel4-be axs101 arcangel4
>    sandbox: +   sandbox
>
> mips is the mipsel libgcc thing, I need to see if eldk5.5 behaves
> better than 5.2.1 in that regard and passing or setting
> CONFIG_USE_PRIVATE_LIBGCC=yes.  sparc is however all warnings (old gcc
> warning when it should not).
>
> Also, is there any way to note which boards/arches an error comes from?
> +/home/trini/work/u-boot/master/.bm-work/00/disk/part.c: In function
> `get_device':
> +/home/trini/work/u-boot/master/.bm-work/00/disk/part.c:453: warning:
> 'hwpart' might be used uninitialized in this function
>
> Is confusing if you hadn't done a sparc solo build before and seen that.

I'll add an option for that.

Regards,
Simon