[mapserver-dev] the state of msautotest

[thomas bonfort]

@Oliver: Thanks for suggesting the travis platform, it's a perfect fit
for what we are testing

@Mike: Thanks for the hosting offer. In the current state, travis-ci
is largely sufficient and better than what we could have setup on our
own in a reasonable amount of time. However providing a remotely
accessible "reference" platform where devs can test their changes and
add expected results might be a usefull tool. Is this something devs
want? Mike, is opening up an ubuntu vm something possible?


Just a quick followup on the rest:

- There's an autotest wiki page in place here:
https://github.com/mapserver/mapserver/wiki/Test-Suite

- The Travis continuous integration is activated and operational for
the 6.2 and master branches. The test suite will be run on any commit
to those branches, as well as on all pull-requests submitted to the
issue tracker (which is another reason to use pull requests rather
than attaching patches, example here:
https://github.com/mapserver/mapserver/pull/4044 )

- I've added a number of autotests that need a postgis db to run on.
They will evidently fail if such a db isn't available. There is
however a script that will automatically create/populate such a db for
you at the root of msautotest. These tests are there to validate some
db specific features (filtering, TIME) as well as provide valgrind
info.

- There's a "make test" target in our mapserver repo, which uses the
binaries from the source directory (i.e. there's no need to make
install before running the test suite). The make target should ensure
that the test suite data is downloaded on first run, however it will
not be updated automatically, which means it is up to you to run "git
submodule update" to catch up with updates to msautotest. The tests
*will* fail if not all compilation options have been activated, c.f.
the "reference platform" on the wiki page.

Please provide any feedback you might have with these new tools.


regards,
thomas