Are you tired of setting up Travis CI or GitHub Actions for all your projects? Then travix
is for you! \o/
- Quickstart
- Building
- Using Travix in your code
- Reasons to use Travix
- Reasons not to use Travix
- How to use git version
Note: Since v0.11.0, Travix only supports Haxe 3.3+. If you need older Haxe, please use v0.10.5.
To use Travix within one of your libs, cd
into your project root and execute:
haxelib install travix # if it's not installed already
haxelib run travix init # this will ask you to input the necessary information and create a .travis.yml or GitHub Actions workflow YAML file
From there, the setup should be straight forward.
Travix has individual commands for building:
interp
- run tests on interpreterneko
- run tests on nekonode
- run tests on nodejs (with hxnodejs)php
- run tests on phpjava
- run tests on javajs
- run tests on puppeteer (headless browser)flash
- run tests on flash (see instructions below)python
- run tests on pythoncs
- run tests on cscpp
- run tests on cpplua
- run tests on luahl
- run tests on hashlink
So instead of having to have to define all kinds of builds and figuring out the right way to run them, this will do.
There are differences among platforms about logging and exiting the process.
For example, we run JS tests on PhantomJS where your test code needs to communicate
with the Phantom host in some special ways in order to log or exit the process.
And on Flash you need to use flash.Lib.trace
and flash.system.System.exit(status)
.
In order to ease the pain, Travix provides a unified interface for these functionalities.
Use them to instead of trace()
, Sys.exit()
, etc, for maximum compatibility across platforms
travix.Logger.print(string)
: Print a string without newlinetravix.Logger.println(string)
: Print a string with newlinetravix.Logger.exit(exitCode)
: Exit the process
If you don't want to introduce a hard compile dependency to Travix in your code for some reason, you can also use the travix.Logger
in combination with the compile condition #if travix
that will result in the travix.Logger
being used when executing builds via
travix but bypass it when executed without.
For example:
inline function println(v:String)
#if travix
travix.Logger.println(v);
#elseif (flash || air || air3)
flash.Lib.trace(v);
#elseif (sys || nodejs)
Sys.println(v);
#else
trace(v);
#end
The BDD library Buddy has built-in support for flash and JS testing, so if you're using Buddy you don't even have to worry about the above.
Apart from helping the pathologically lazy to set up a CI, the strength of Travix lies in that it deals with dependencies rather gracefully:
- it relies on the
haxelib.json
to install haxelib dependencies. It also uses thehaxelib dev
command to "mount" your library as a haxelib, giving you all the extra features, e.g. the presense of your-D libname
flag and the inclusion ofextraParams.hxml
in the build. This happens with theinstall
command. - it follows a fail-fast philosophy. What's that supposed to mean? Normally, in your CI, you will install all dependencies before running any of the tests. If you wait for the installation of hxjava, hxcpp, hxcs, mono and php, only to make your first test abort because of a missing semi-colon or a similarly silly mistake, it can be rather frustrating. To avoid that problem, Travix diverges from the usual modus operandi of having distinct installation and execution phases, and instead installs such dependencies right before execution, e.g. in the
cs
command.
The motivation behind Travix is to be able to spin up CI setups quickly, for many small libraries (in my case the tink
libs). It is very likely, that it will not scale up to bigger projects, particularly when multiple builds need to be run in unison to have a test. If you have suggestions - or better yet: pull requests - to make Travix more useful for such cases, you are highly welcome.
In your .travis.yml
simply replace haxelib install travix
with the following:
haxelib git travix https://github.com/back2dos/travix && pushd . && cd $(haxelib config)travix/git && haxe build-neko.hxml && popd