Re: Anyone working on testbed barriers?

[Christian Grothoff]

On 1/11/21 2:18 PM, TheJackiMonster wrote:
> Hi,
> 
> today I got my parts of my messenger testcase running with multiple
> peers but I encountered a problem with the barriers. The
> GNUNET_TESTBED_barrier_wait() call checks if it gets called from inside
> of a testbed by checking enviroment variables which do net get set
> automatically when you start a testbed via GNUNET_TESTBED_test_run().
> 
> I figured out how to quickfix this by saving the old value of the
> specific environment variable ( "GNUNET_TESTBED_CONTROLLER_CONFIG" )
> before starting the testbed and resetting it to its original value
> after ending the testbed. But I'm not sure if this should be solved
> like this. Other testbed functions do not use this environment variable
> and some of them even use the default configuration.
> 
> Should this be fixed differently? Otherwise I would prepare a commit to
> master with my solution.
> 
> There is also another problem with the barriers now. After calling
> GNUNET_TESTBED_barrier_wait() I get a segfault. I can probably look
> into that to fix it but maybe someone else has more insight what could
> be wrong. Has anyone used testbed barriers before? ^^'

Sree Harsha did, a long time ago, but the code has not been maintained /
used for many years, so it likely needs some major work to function
properly.

Alas, TNG will _also_ imply major changes to testbed, so IMO quickfixes
are fine for now, and bigger work should happen after TNG has stabilized
and we begin testing everything with TNG.

-Christian