Re: [PATCH] xen/arm: Fix arm32 build failure when early printk is enabled

[Michal Orzel <michal.orzel@xxxxxxx>]

On 29/02/2024 11:10, Julien Grall wrote:
> 
> 
> Hi,
> 
> On 29/02/2024 10:07, Michal Orzel wrote:
>>
>>
>> On 28/02/2024 23:27, Stefano Stabellini wrote:
>>>
>>>
>>> On Wed, 28 Feb 2024, Michal Orzel wrote:
>>>> Hi Julien,
>>>>
>>>> On 28/02/2024 12:42, Julien Grall wrote:
>>>>>
>>>>>
>>>>> Hi Michal,
>>>>>
>>>>> On 28/02/2024 10:35, Michal Orzel wrote:
>>>>>> Commit 0441c3acc7e9 forgot to rename FIXMAP_CONSOLE to FIX_CONSOLE in
>>>>>> TEMPORARY_EARLY_UART_VIRTUAL_ADDRESS macro. This results in a build
>>>>>> failure on arm32, when early printk is enabled:
>>>>>> arch/arm/arm32/mmu/head.S:311: Error: invalid operands (*UND* and *ABS* 
>>>>>> sections) for `*'
>>>>>
>>>>> Good catch! Somewhat related I wonder whether we should add earlyprintk
>>>>> testing in gitlab?
>>>> I thought about adding this and I think we should at least have build jobs 
>>>> (hypervisor only, no toolstack)
>>>> selecting early printk. When it comes to testing if early printk works, 
>>>> I'm not sure. It'd be nice
>>>> but FWIR we have limited bandwidth.
>>>>
>>>> @Stefano, what's your opinion?
>>>
>>> I think it would be a good and quick test to have. To save testing
>>> bandwidth I think we should reduce the amount of debug/non-debug
>>> variations of the same tests that we have.
>> Yes, I suggested that some time ago. We could keep both versions for generic 
>> tests,
>> but remove the non-debug version (unless you prefer to do the opposite) for:
> 
> I think it makes sense during development window to use the debug
> version. However, I think we want some non-debug testing during the
> hardening phase.
> 
> Can gitlab read CONFIG_DEBUG from Config.mk?
At the moment, we have 2 types of jobs - non debug and debug (with -debug 
suffix).
They set "debug" variable accordingly, which is used later on to modify .config:
echo "CONFIG_DEBUG=${debug}" >> xen/.config

Without this line, Xen would be built according to default value of 
CONFIG_DEBUG.
That said, I don't think we want to get back to this behavior.

If we want to save some bandwidth, we should make a decision whether to keep 
debug or non-debug versions.
x86 has both versions for build jobs and mostly debug test jobs.

~Michal