Do not fail when gen_protos cannot be imported

[aaltay]

R: @robertwb

[coveralls]

Coverage increased (+0.04%) to 70.698% when pulling 930a1e3 on aaltay:lnt into 63bcd39 on apache:master.

[robertwb]

When would we expect this to happen?

[asfbot]

Build finished.
--none--

[aaltay]

It happens, if the sdk is packaged without the egg_info. In that case pip install, triggers python setup.py egg_info. Even though this cmd does not requires protos, it fails because packaged SDK does not contain the gen_protos.py.

[robertwb]

How about in that case wrapping the commands with classes that raise an error? This way we never silently fail to generate the protos when we need them.

[robertwb]

Alternatively, move the import into the run method.

[aaltay]

Would not it be the same problem, if we move the import into the run method? Then it will fail when run is called.

[robertwb]

But in that case it would only fail for the commands that require the generated protos. egg_info would be just fine.

…

On Fri, Jun 2, 2017 at 10:39 AM, Ahmet Altay ***@***.***> wrote: Would not it be the same problem, if we move the import into the run method? Then it will fail when run is called. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3279 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAdqgZXGjMrB3vmV-7rJtddE7_3udcT1ks5sAEjOgaJpZM4NtXob> .

[aaltay]

@robertwb Maybe I am missing something, but moving the import into the run method fails the same way. Because egg_info, triggers running of build_py. (That was the reason for original failure, egg_info did indeed caused a call to generate_protos_first and that returns a wrapped cmd.)

[robertwb]

Ah, I wasn't aware that egg_info called build_py.

LGTM.

[aaltay]

Thank you, merging.