Re: [gpsd-dev] [PATCH 4/4] Parallelizes gps-regress and gps-makeregress.

[Robert Norris]

----------------------------------------
> Date: Mon, 29 Feb 2016 14:13:31 -0800
> From: address@hidden
> To: address@hidden
> Subject: Re: [gpsd-dev] [PATCH 4/4] Parallelizes gps-regress and 
> gps-makeregress.
>
>
>> patch set merged, looks good!
>
> Thanks, though note that one of my earlier as yet uncommitted patches is
> needed to make this maximally useful. Running 'check' in parallel doesn't
> work due to the implied 'testclean' stomping on other tests while they're
> running, hence the new 'test-noclean'. Though running 'gps-regress' in
> parallel by itself works fine without the other patch.

I think it's better to simply remove the use of testclean within the check 
stage as it seems pointless (remove the testclean completely  and also 
'test_spec' within the check stage as the time_regress will automatically build 
it if necessary).

I don't understand why these where added and anyway the clean stage removes all 
generated output (i.e. including the test programs).

Instead I was going to post this simple patch:

diff --git a/SConstruct b/SConstruct
index ce64983..8058925 100644
@@ -1757,6 +1757,7 @@ testclean = Utility('testclean', [],
 check = env.Alias('check', [
     describe,
     python_compilation_regress,
+    testprogs,
     method_regress,
     bits_regress,
     matrix_regress,

Thus this ensures *all* test programs are at least compiled (and should have 
flagged the failing test_gpsmm earlier).

Not all test programs are actually automatically run as part of tests,
 (which only builds the needed programs) thus ensuring they all compile.

Presumably this would also be beneficial for running regression tests in 
parallel.

--
Be Seeing You - Rob.
If at first you don't succeed,
then skydiving isn't for you.