On 4/20/21 9:01 AM, Coly Li wrote:
> On 4/20/21 9:36 PM, Jens Axboe wrote:
>> On 4/19/21 10:20 PM, Coly Li wrote:
>>> On 4/20/21 12:02 PM, kernel test robot wrote:
>>>> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git for-5.13/drivers
>>>> head:   637d8f534626747ad165c9d28c6673d88d39399d
>>>> commit: 97d69b16cb973e04ea5309b9cb4356aa6b42c54e [44/80] bcache: add initial data structures for nvm pages
>>>> config: i386-randconfig-a012-20210419 (attached as .config)
>>>> compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
>>>> reproduce (this is a W=1 build):
>>>>         # https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git/commit/?id=97d69b16cb973e04ea5309b9cb4356aa6b42c54e
>>>>         git remote add block https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git
>>>>         git fetch --no-tags block for-5.13/drivers
>>>>         git checkout 97d69b16cb973e04ea5309b9cb4356aa6b42c54e
>>>>         # save the attached .config to linux build tree
>>>>         make W=1 W=1 ARCH=i386 
>>>>
>>>> If you fix the issue, kindly add following tag as appropriate
>>>> Reported-by: kernel test robot <lkp@intel.com>
>>>>
>>>> All errors (new ones prefixed by >>):
>>>>
>>>>    In file included from <command-line>:32:
>>>>>> ./usr/include/linux/bcache-nvm.h:109:3: error: #error "Non-64bit platform is not supported"
>>>>      109 |  #error "Non-64bit platform is not supported"
>>>>          |   ^~~~~
>>>>
>>>
>>> This series is result of community cooperation, "#error" is to make sure
>>> this header won't be used in non-64bit platform. And following patches
>>> in this series make sure the "#error" won't be triggered.
>>
>> It sounds like the series is ordered wrong, then. You don't introduce a
>> known failure point, then fix it up in subsequent patches. You ensure
>> that the condition can never happen, then introduce the change.
>> Otherwise you're left with broken bisection points, which is
>> problematic.
>>
>> I need to take a closer look, but will any 32-bit kernel compile fail at
>> that commit???
>>
> 
> The whole series won't fail 32-bit kernel compiling. But, don't spend
> you time on it for now, I will prepare a better code base for 5.14 merge
> window.

Right, but in the middle of it it will? That's kind of the point. We
can't have known breakage in the middle of a series, it'll ruin any
attempts at bisecting issues on 32-bit if we happen to end up in the
broken area at some point.

-- 
Jens Axboe