On Sun, Jun 23, 2013 at 1:37 PM, kristian <m.krist...@web.de> wrote:
>> * Mavenizing.
> hey - I am more than happy to help with that task if help is needed ;)
We welcome any help :-) I think we really want to do this in master
*before* branching, since it will mean structural changes that
complicate merging fixes from 9k.
Difficult bits of mavenizing, as I see them:
* Libraries we control and iterate (sometimes frequently) need a
snapshot repository. All of those projects need trivial processes for
pushing new snapshots, so we can always build JRuby with them.
A quick list, scanning build_lib...
Already being published to Maven for releases (but not snapshots):
bytelist
invokebinder
jcodings
jffi
jnr-constants
jnr-enxio
jnr-ffi
jnr-netdb
jnr-posix
jnr-unixsocket
jnr-x86asm
joni
Not in maven, but need to be:
coro-mock (compile only)
jsr292-mock (compile only)
unsafe-mock (compile-only)
yecht (1.8 yaml library...going away in 9k, but still needed in 1.7.x)
yydebug
* Bootstrapping, build and test flow.
Remaining test suites in ant need to move to Rake so we can trigger
them from there. Maven would then be used *solely* to (re)build JRuby,
or to run a "complete" test suite (by calling out to Rake, ideally).
The Rake targets would depend on jruby.jar, which would be built by
maven. So the build process would probably become "build JRuby via
maven, run rake tasks for testing" and ideally after the first build,
you'd only need to run Rake.
* Transitioning
I'd really like to see this happen as soon as possible so we can work
out kinks long before the 1.7.5 release. Neither Tom nor I (nor most
of the other core folks) are particularly thrilled about switching to
Maven for builds, but there's so many problems it solves. I think most
everyone's on board to make the change.
Start a branch and we'll see what we can accomplish?
- Charlie
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
