unit-test with TavisCI about haxelib HOT 12 CLOSED

I think this would be a brilliant idea.

from haxelib.

I think before we start using travis we would need an appropriate amount of tests to run. These tests need to be written first.

Also I think that the current version of this repository only compiles with an SVN version of Haxe.

I am not generally against it, but I think there are a few steps that should be done before.

from haxelib.

Well I suppose it can hardly do any harm, so feel free to set it up as you wish. But I agree with Hippo that it might be a little early on for this now.

Does TravisCI also support Mac and Windows? One way or another we need to cover those platforms. If TravisCI doesn't do that, then we will need a Haxe-based solution for Windows and Mac and I see no point in duplicating the maintenance overhead in Linux.

from haxelib.

Currently travis-ci is Linux only. (Travis issue) (Vagrant, which is is the underlying technology of Travis (and relies heavily on VirtualBox headless) very recently got VM-Ware support so technically it is possible right now, but I don't expect this to happen very soon.) So is isn't 100% suitable for this project.

P.S.: I would like to see a test suit for haxelib that at least could be run locally.

from haxelib.

In fact my idea is to create a unit test runner written in haxe, which can be run locally, and TravisCI will help us to run it on its Linux VM. Of course we should also run the test ourself in Mac/Windows machines. And of course we need to create the unit test before using TravisCI.

The benefit of using TravisCI is an extra level of safety that each commit and pull request is unit-tested at least once.

So I guess we actually all agree on this :)

@TheHippo Can you point me to which specific part or issue that requires an SVN version of haxe? We can already install haxe3RC on TravisCI, and should be able to further upgrade it using some simple shell script.

from haxelib.

@andyli #7

from haxelib.

I suppose it wouldn't affect the unit-test running on TravisCI since we are very unlikely to use compilation server for it.

from haxelib.

That's right. No one want to do not use TravisCI here. In my opinions this is just not high priority right now, because:

• We have 0% test coverage
• (Tests would be only run on Linux.)

When some tests are written it would be cool to have some CI. It is indeed very helpful to see if a pull request will merge cleanly and do not break any tests.

from haxelib.

Hi Andy

Rather than creating a new unit testing server is there an ability to tie
in with one of the existing frameworks? I know munit has a while
infrastructure for running tests locally and generating reports in a custom
format for a CI server, so that might be a good start. Not sure if utest or
other frameworks have anything similar?

Worth checking in case you don't have to reinvent the wheel...

Jason
On Mar 20, 2013 3:02 AM, "Andy Li" [email protected] wrote:

In fact my idea is to create a unit test runner written in haxe, which can
be run locally, and TravisCI will help us to run it on its Linux VM. Of
course we should also run the test ourself in Mac/Windows machines. And of
course we need to create the unit test before using TravisCI.

The benefit of using TravisCI is an extra level of safety that each commit
and pull request is unit-tested at least once.

So I guess we actually all agree on this :

@TheHippo https://github.com/TheHippo Can you point me to which
specific part or issue that requires an SVN version of haxe? We can already
install haxe3RC on TravisCI, and should be able to further upgrade it using
some simple shell script.

—
Reply to this email directly or view it on GitHubhttps://github.com/back2dos/haxelib/issues/19#issuecomment-15136033
.

from haxelib.

Well, I'm not sure this would really mean to "reinvent the wheel". To my knowledge NME also uses Travis-CI. There's nothing really wrong with it. But as Hippo pointed out, right now we should really write the tests to start with ;)

from haxelib.

btw, I don't think there will be a lot of windows-specific code, so it seems to me that a linux CI server would be fine

from haxelib.

Haxelib is "on Travis" now: https://travis-ci.org/HaxeFoundation/haxelib

from haxelib.