Re: [Qemu-devel] [PATCH] blogposts: add post about the new check-tcg infrastructure

[Max Filippov]

On Fri, Jun 22, 2018 at 2:03 AM, Alex Bennée <address@hidden> wrote:
>
> Max Filippov <address@hidden> writes:
>
>> On Thu, Jun 21, 2018 at 11:41 AM, Alex Bennée <address@hidden> wrote:
>>> Signed-off-by: Alex Bennée <address@hidden>
>>> ---
>>>  _posts/2018-06-21-tcg-testing.md | 129 +++++++++++++++++++++++++++++++
>>>  1 file changed, 129 insertions(+)
>>>  create mode 100644 _posts/2018-06-21-tcg-testing.md
>>>
>>> diff --git a/_posts/2018-06-21-tcg-testing.md 
>>> b/_posts/2018-06-21-tcg-testing.md
>>
>> [...]
>>
>>> +The `tests/tcg` directory still contains a number of source files we
>>> +don't build. Notably the cris, lm32, mips, openrisc and xtensa targets have
>>> +a set of tests that need a system emulator. Now we have the
>>> +infrastructure for compiling I hope we can get support for system
>>> +tests added fairly quickly. There will need to be some work to figure
>>> +out a nice common way to pass results back to the build-system. For
>>> +linux-user this is simple as all programs can simply return their exit
>>> +code however for system emulation this is a little more involved.
>>
>> xtensa tests pass exit codes to the build system through semihosting calls.
>> If any of them fails make check fails as well.
>
> I've re-written that section as:
>
>   The `tests/tcg` directory still contains a number of source files we
>   don't build.
>
>   The cris and openrisc directories contain user-space tests which just
>   need the support of a toolchain and the relevant Makefile plumbing to
>   be added.
>
>   The lm32, mips and xtensa targets have a set of tests that need a
>   system emulator. Aside from adding the compilers as docker images some
>   additional work is needed to handle the differences between plain
>   linux-user tests which can simply return an exit code to getting the
>   results from a qemu-system emulation. Some architectures have
>   semi-hosting support already for this while others report their test
>   status over a simple serial link which will need to be parsed and
>   handled in the `run-%:` test rule.
>
> How is that?

'run-%' goal is only present in xtensa Makefile, other test suites have
explicit loop in the 'check' goal. Otherwise LGTM.

> Any chance you could look into what it would take to package up the
> xtensa toolchain in a docker container?

Can you point me to an example?

> Are they simply tarballs of binaries?

Yes, we have that option:

  https://github.com/foss-xtensa/toolchain/releases

Or they may be built from source.

-- 
Thanks.
-- Max