Re: [gpsd-dev] regression test puzling non-failure which should probably fail

[Greg Troxel]

I see, so this is a failure to run the test, not a test failure.  Is it
really trying the port in one half before forking?  I am seeing it
fairly often, 3/13 on my netsbd-5 adm64 box (which does a lot of other
stuff).

I wonder if it would make sense to try running the test up to 10 times,
to avoid losing test coverage.

I'm also seeing this (on mac):

gpsfake: socket error [Errno 57] Socket is not connected.

pgpVqtoByOhz3.pgp
Description: PGP signature