GCC version for RTEMS 6?

[Chris Johns]

On 6/5/2022 4:16 pm, Sebastian Huber wrote:
> Hello Chris,
> 
> thanks for the tests.

My pleasure. Thank your gcc12.

> On 05/05/2022 11:28, Chris Johns wrote:
>> On 4/5/2022 3:54 pm, Sebastian Huber wrote:
>>> On 04/05/2022 02:14, Chris Johns wrote:
>>>> On 3/5/2022 4:56 pm, Sebastian Huber wrote:
>>>> Are there any posted test results?
>>>
>>> I did only local test runs on simulators so far.
>>>
>>
>> I have posted some test results for the erc32 on SIS, zynq on qemu and
>> ultrascale on hardware:
>>
>> https://lists.rtems.org/pipermail/build/2022-May/033152.html
>> https://lists.rtems.org/pipermail/build/2022-May/033153.html
>> https://lists.rtems.org/pipermail/build/2022-May/033154.html
>>
>> This covers SPARC on a simulator, ARM 32 bit on qemu and ARM 64bit on hardware.
>>
>> - There are some new warnings which is expected
>>
>> - erc32
>>
>> It has 3 new failures compared to the test results posted by Joel earlier this
>> month. They are:
>>
>>      crypt01.exe
> 
> The crypt01 is a long running test.

OK.

>>      spintrcritical23.exe
> 
> I see also sporadic failures in the spintrcritical* tests on this BSP
> independent of the compiler.
> 
>>      minimum.exe
> 
> The minimum test is an issue on all BSPs since the RTEMS tester has no way to
> figure out what to do since the test has no output. One option would be to use
> ELF files for the tester input with ELF notes for the test state, etc.

This is a fantastic is idea. We can also embed more information as we need
relating to load considerations, time outs and more.

> 
>>
>> - arm
>>
>> This seems fine. I do not have a base line to compare. The tester seemed to
>> crash if the log and log mode options are not specified.
>>
>> - aarch64 bsps
>>
>> These BSPs generated lots of warnings in the testsuite due to this line:
>>
>>   https://git.rtems.org/rtems/tree/bsps/aarch64/include/bsp/start.h#n175
>>
>> The warning is about comparing arrays. I have not looked deeper.
> 
> This can be fixed by using RTEMS_OBFUSCATE_VARIABLE().

Ah ok. Thanks.

>>
>> The aarch64 test results are very good. I have not tested the Versal (A72) but I
>> will tomorrow test libbsd on that that architecture. There are 3 failures:
>>
>>   malloc04.exe
>>   psxconfig01.exe
>>   ts-validation-no-clock-0.exe
>>
>> The only one I am not sure about it the last one.
>>
>> Conclusion:
>>
>> I think the results show we can move to gcc12. There are some rough edges
>> related to warnings we need to clean up and I am sure they will addressed.
> 
> Ok, good. Lets wait for the GCC 12.1 release.
> 

Sure.

>>
>> Sebastian, if you wish to make the change I am ok but I ask if you are OK with
>> the testing solution I posted could you please push that patch and make the
>> change on it?
> 
> Which patch and change are you referring to?
> 

The patch you posted is now not valid as I have changed the default files. It is
nothing more than that.

Chris