Output appropriate error message on help failure and minor cleanups

[chu11]

No description provided.

[coveralls]

Coverage decreased (-0.06%) to 75.754% when pulling e8f81cc on chu11:errordocsnotbuilt into 4801f62 on flux-framework:master.

[coveralls]

Coverage decreased (-0.06%) to 75.75% when pulling 896d22f on chu11:errordocsnotbuilt into 4801f62 on flux-framework:master.

[chu11]

Hmmm, lots of failures in make check, it appears a2x can't be found in travis

checking for a2x... no

probably have to add asciidoc to travis.yml. But before I do that, I don't see what could have happened between #871 and the current master.

[grondo]

We purposefully leave a2x out of Travis, it takes too long to install all
its dependencies. Why does make check fail without a2x now?

On Oct 30, 2016 3:41 PM, "Al Chu" notifications@github.com wrote:

Hmmm, lots of failures in make check, it appears a2x can't be found in
travis

checking for a2x... no

probably have to add asciidoc to travis.yml. But before I do that, I don't
see what could have happened between #871
#871 and the current
master.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#887 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAtSUp74vgduDxWjqCPRM-Y67zQ5LIcCks5q5R0TgaJpZM4KkZuj
.

[grondo]

Ah, the problem is (at the least) with the flux-help tests in t0001-basic.t. These tests use fake manpages to ensure flux help properly invokes man even when manpages aren't built, so the .nodocs file breaks these tests.

A possible workaround would be some kind of environment variable to disable the .nodocs test, or possibly point flux-help to a different config dir without the .nodocs file, etc... Then use that for any tests exercising flux-help

[chu11]

Ahhh. I realized it was fake manpages in the tests, but I didn't quite get how things worked with a2x not being installed. I now see with the commit 0e13c83, the logic surrounding a2x being required/not required changed.

Yeah, I'll have to tweak the tests in some way.

[grondo]

I guess it would be safe if you removed .nodocs for the duration of those tests, if nothing else in make check is dependent on the .nodocs file (keep in mind someday parallel-tests might work, so multiple tests could be running at the same time)

[chu11]

I elected to support a FLUX_IGNORE_NO_DOCS environment variable as a workaround. It's not the cleanest solution, but I think it ensures the unit tests will work regardless if docs are generated or not, and should be safe for any future parallelism in the testsuite in the future.

[coveralls]

Coverage decreased (-0.04%) to 75.771% when pulling 0030676 on chu11:errordocsnotbuilt into 4801f62 on flux-framework:master.

[codecov-io]

Current coverage is 72.15% (diff: 66.66%)

Merging #887 into master will decrease coverage by 0.06%

@@             master       #887   diff @@
==========================================
  Files           156        156          
  Lines         26929      26935     +6   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
- Hits          19448      19436    -12   
- Misses         7481       7499    +18   
  Partials          0          0          

Diff Coverage	File Path
•••••• 66%	src/cmd/builtin/help.c
•••••••••• 100%	src/common/libflux/conf.c

Powered by Codecov. Last update 4801f62...0030676

[chu11]

because of the workaround environment variable, code coverage does go down a tad as some of the new .nodocs code won't be touched.

[grondo]

Yeah, seems fine to me. Merging this one thanks!