Re: [Qemu-devel] [RFC PATCH] configure: Enable out-of-tree acceptance tests

[Cleber Rosa]

On 06/06/2018 04:11 PM, Eduardo Habkost wrote:
> On Wed, Jun 06, 2018 at 04:36:16PM -0300, Philippe Mathieu-Daudé wrote:
>> On 06/06/2018 04:24 PM, Eduardo Habkost wrote:
>>> On Tue, Jun 05, 2018 at 11:45:03AM -0300, Philippe Mathieu-Daudé wrote:
>>> [... something about config files ...]
>>>>> And after that, the following would run all "console" tests:
>>>>>
>>>>>   avocado run -t console
>>>>>
>>>>> How does this sound?
>>>>
>>>> For my use cases this doesn't worry me, I'll let Eduardo/Fam opine about
>>>> use_test_dir_when_no_references_given.
>>>
>>> Well, I can't give an opinion because I couldn't understand
>>> what's the final goal here.
>>
>> You cut too much, the relevant part is:
>>
>>   On 05/30/2018 10:06 PM, Cleber Rosa wrote:
>>   > But at the same, there are security implications: `list` won't
>>   > load/execute any test code (different from, say, standard Python
>>   > unittests), while "run" obviously will.  So "avocado run" may end up
>>   > running what users don't want if a malicious user controls
>>   > "$avocado_datadir_paths_test_dir".
> 
> The final goal still isn't clear to me.  Why would somebody want
> to use $avocado_datadir_paths_test_dir instead of just specifying
> "." in the command-line?
> 
> 
>>
>>>
>>> What exactly is missing in the current solution?  Why
>>> "avocado run -t console ." wouldn't work?
>>
>> It doesn't work in out-of-tree builds, I have to use the full path:
>>
>>   >>   build_dir$ avocado run
>> /full/path/to/sources/qemu/tests/acceptance/boot_linux_console.py
> 
> Isn't the symlink you suggested a better solution than requiring
> the user to edit a config file?
> 

It's definitely the simplest, and would avoid a questionable knob and
behavior in Avocado.

- Cleber.