Add console command that prints needed envvars

[allardhoeve]

When running nosetests or python manage.py test there are often left-over processes that are never reaped. As I use watch to run the tests every tenth of a second, that results in many, many processes 😄

I've traced this back to the re-exec that is performed. I would like to avoid that re-exec cleanly, but currently I have to have read the libfaketime source code to find out which envvars it wants. And keep the vars in sync ofcourse.

So, let's add a console script that prints out the needed envvars, like ssh-agent does on start. Then you can call libfaketime and it will output:

export LD_PRELOAD="/home/.../vendor/libfaketime/src/libfaketime.so.1"
export FAKETIME_DID_REEXEC=true

This can then be used at startup:

eval $(libfaketime)
nosetests

[allardhoeve]

@gertjanol

[simon-weber]

Huh, any idea where the orphan processes are coming from? We can track that separately if you'd like; making the non-hack way easier is a fine workaround.

One quick suggestion: can we make the top-level executable python-libfaketime instead? I'm worried about conflicting with the original c library.

[allardhoeve]

No, I have no idea where they are coming from. And I do not really have the time to check it out. It might have something to do with the fact that we run 1k tests or so and use all CPU cores to run them. The nosetests multi process plugin is... not very nice.

But having a nice supported way of circumventing the respawn seems nice in any case.

[allardhoeve]

Also, updated!

[allardhoeve]

And finally, some bonus documentation 😄

[simon-weber]

Looks great; thanks!

[allardhoeve]

Cool. If you upload to pypi, can you give me a ping?

[simon-weber]

Sure, I'll try to do that tonight.

[simon-weber]

@allardhoeve ok, 0.3.0 is up!

[allardhoeve]

Thanks! Works like a charm!