[yocto] Kernel patch is unpacked but not applied

[Bruce Ashfield]

On 13-06-27 08:50 AM, Bryan Evenson wrote:
>> -----Original Message-----
>> From: Bruce Ashfield [mailto:bruce.ashfield at windriver.com]
>> Sent: Thursday, June 27, 2013 1:24 AM
>> To: Bryan Evenson
>> Cc: yocto at yoctoproject.org
>> Subject: Re: [yocto] Kernel patch is unpacked but not applied
>>
>> On 13-06-26 11:42 PM, Bruce Ashfield wrote:
>>> On 13-06-26 11:08 PM, Bryan Evenson wrote:
>>>> I am building a custom Linux kernel under poky-dylan and I am having
>>>> an issue with patches being unpacked but not applied.  I have two
>>>> layers that have a bbappend file for the kernel.  Generally, this is
>>>> what each bbappend file looks like:
>>>>
>>>> #First .bbappend
>>>>
>>>> FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:"
>>>> COMPATIBLE_MACHINE_mach1 = "mach1"
>>>> COMPATIBLE_MACHINE_mach2 = "mach2"
>>>> SRC_URI_append_mach2 = " file://${MACHINE}/${KBRANCH}/0001-
>> blah.patch \
>>>>       file://${MACHINE}/${KBRANCH}/0002-blah.patch \
>>>>       "
>>>>
>>>> # Increment the recipe revision
>>>> PRINC := "${@int(PRINC) + 1}"
>>>>
>>>> # Second .bbappend
>>>>
>>>> FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:"
>>>> COMPATIBLE_MACHINE_mach1 = "mach1"
>>>> COMPATIBLE_MACHINE_mach2 = "mach2"
>>>> SRC_URI_append_mach2 = " file://${MACHINE}/${KBRANCH}/0003-
>> blah.patch \
>>>>       file://${MACHINE}/${KBRANCH}/0004-blah.patch \
>>>>       "
>>>>
>>>> # Increment the recipe revision
>>>> PRINC := "${@int(PRINC) + 1}"
>>>>
>>>> All of the patch files are properly unpacked, but the last two
>>>> patches are not being applied.  If I open the devshell for the
>> kernel
>>>> (bitbake -c devshell linux-yocto-custom) I can see in the git log
>>>> that the first patch set was applied.  The second patch set exists
>>>> but is not applied.  If I call "guilt push" repeatedly from the
>>>> devshell, the patches from the second set are cleanly applied.  My
>>>> guess is that something doesn't like the second SRC_URI_append call.
>>>> Any ideas on how to fix it?
>>>
>>> It shouldn't matter. Let me try and set up something that reproduces
>>> the problem and get back to you.
>>>
>>> Strangely .. I'm actively debugging something similar here already,
>> so
>>> I may be able to re-use it.
>>>
>>> Stay tuned.
>>>
>>> Out of curiosity, have you tried this on master ?
>>
>
> I have not tried this on master.  Here is the relevant build information:
>
> Build Configuration:
> BB_VERSION        = "1.18.0"
> BUILD_SYS         = "i686-linux"
> NATIVELSBSTRING   = "Ubuntu-12.04"
> TARGET_SYS        = "arm-poky-linux-gnueabi"
> MACHINE           = "at91sam9x5ek"
> DISTRO            = "poky"
> DISTRO_VERSION    = "1.4.1"
> TUNE_FEATURES     = "armv5 thumb dsp arm926ejs"
> TARGET_FPU        = "soft"
> meta-melink       = "dylan:e4a0a4c5e419154a34710a1b6c28d4180c6304c3"
> meta-atmel        = "master:b2a9c072730b0f6b2a669a8de613e5e1d59453e6"
> meta
> meta-yocto
> meta-yocto-bsp    = "dylan:e4a0a4c5e419154a34710a1b6c28d4180c6304c3"
> meta-oe           = "dylan:13ae5105ee30410136beeae66ec41ee4a8a2e2b0"
>
> So I'm at the HEAD of the Dylan branch for both poky and meta-oe.
> The kernel patches are in both the meta-atmel layer and the meta-melink
> Layter.  Ignore that version string for meta-melink; it's not a git
> repository so it's just grabbing the info from poky.
>
>> So I tried to recreate this on master, and things worked for me. Which
>> I figured would happen, since this will make it harder to fix ... and
>> that's how it always works out.
>>
>> I created two layers, with four patches to the main linux Makefile.
>> They all add something to the description:
>>
>> FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:"
>>
>> SRC_URI_append = " file://0001-makefile-one.patch \
>>                      file://0002-makefile-two.patch"
>>
>>
>>
>> PRINC := "${@int(PRINC) + 1}"
>>
>> and
>>
>>
>> FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:"
>> SRC_URI_append = " file://0003-makefile-three.patch \
>>                      file://0004-makefile-four.patch"
>>
>>
>>
>> PRINC := "${@int(PRINC) + 1}"
>>
>> .. and alas, the all were applied:
>>
>>   > grep NAME Makefile
>> NAME = Displaced Humerus Anterior one two three four
>>
>> ....
>>
>> Can you send me the exact names of your patches ? I'm wondering if an
>> already applied check is triggering and preventing the auto push.
>>
>
> Here are the patch names in the first layer (meta-atmel):
>
> " file://${MACHINE}/${KBRANCH}/UBI.cfg \
>      file://${MACHINE}/${KBRANCH}/dma.cfg \
>      file://${MACHINE}/${KBRANCH}/usb-c.patch \
>      file://${MACHINE}/${KBRANCH}/usart3.patch \
>      file://${MACHINE}/${KBRANCH}/rtc.patch \
>      file://${MACHINE}/${KBRANCH}/serial_dma.patch \
>      ${@base_contains("MACHINE_FEATURES", "watchdog", "file://${MACHINE}/${KBRANCH}/watchdog.patch", " ", d)} \
>      "
>
> If you're wondering about that last line, I added a watchdog
> feature to the meta-atmel layer to enable or disable the watchdog
> for the kernel, U-Boot and the AT91Bootstrap.
>
> And here are the patch names for the second layer (meta-melink):
>
> SRC_URI_append_at91sam9x5ek = " file://${MACHINE}/${KBRANCH}/mq.cfg \
>      file://${MACHINE}/${KBRANCH}/gpio.cfg \
>      file://${MACHINE}/${KBRANCH}/ad5446.cfg \
>      file://${MACHINE}/${KBRANCH}/rs485.patch;apply=yes \
>      file://${MACHINE}/${KBRANCH}/usart1.patch;apply=yes \
>      file://${MACHINE}/${KBRANCH}/mmc.patch;apply=yes \
>      file://${MACHINE}/${KBRANCH}/one-wire.patch;apply=yes \
>      "
>
> I tried adding the "apply=yes" to the second layer patches, and they still

That wouldn't have made a difference, the flags to patches are .. a bit
arcane at times. If a patch is in a SRC_URI it should be applied, and
won't be ignored. So there's a bug lurking somewhere that is specific
to your config.

Out of curiosity, can you try creating a .scc file, containing the
patches and .cfg files within it, and only listing the .scc file on the
SRC_URI.

Your method should work, but I'm trying to keep you moving while I
poke around, and .scc files are specifically designed to help manage
patch lists like this.

To do it, just create "foo.scc" in the same directory as your patches
and configs, and list foo.scc on your SRC_URI .. and that's it.

In foo.scc:

   patch rs485.patch
   patch usart1.patch
   patch mmc.patch
   one-wire.patch

   kconf hardware gpio.cfg
   kconf hardware ad5446.cfg


What I have above is generated to wrap the patches and .cfgs, which allows
them to be audited, and handled the same as in-tree kernel features.
I suspect something is just happening with the wrapping.

> were not applied.  Incidentally, the configuration changes (.cfg files)
> *are* being applied.  So if I understand correctly, do_unpack is getting
> the correct files, do_patch is not applying all the patches, but
> do_configure is applying all the configuration changes.
>
> If it helps, the meta-atmel layer I am using is available here:
> https://github.com/evensonbryan/meta-atmel.  I just pushed an update a few
> minutes ago, so it has what I am using.

Great. If I can use your exact configuration and layers, that makes
it much easier. If I can reproduce the issue, I can fix it.

Bruce

>
> I'll let you know if I discover anything new.
>
> Thanks,
> Bryan
>
>> Bruce
>>
>>>
>>> Cheers,
>>>
>>> Bruce
>>>
>>>>
>>>> Thanks,
>>>> Bryan
>>>> _______________________________________________
>>>> yocto mailing list
>>>> yocto at yoctoproject.org
>>>> https://lists.yoctoproject.org/listinfo/yocto
>>>>
>>>
>>> _______________________________________________
>>> yocto mailing list
>>> yocto at yoctoproject.org
>>> https://lists.yoctoproject.org/listinfo/yocto
>