Christian Edward Gruber wrote:
Yeah, I sort of meant the "local repo server" idea to be ironic, but
my deadpan didn't quite translate into e-mail. <sigh> Anyway, yes,
I'd just like a simple locking mechanism in 2.1 (or heck, in 2.0.x,
since it shouldn't change any behaviour/config except to cover an
edge-case not currently supported in Maven anyway).
Actually - this idea saves a lot of trouble, but also presents some.
Imagine a resident artifact resolver that anticipates your dependencies
and pre-fetches them (maybe talking to other resolvers). Unfortunately
it presents a lot of problems, and security is not the last one :(
I am adding simple file-based locking to Mercury's LocalRepositoryWriter
and also parallelize the writer for two reasons: optimization plus
[proper] unit test automatically addresses mutli-threaded access use case.
BTW - java.nio FileLock did not work for me under Mac OS 10.5, Java
1.5.0_13 - it allowed multiple threads obtain exclusive lock on the same
FileChannel. Does anyone have positive experience with java.nio locking?
It's supposed to be OS bound and available across processes, that serves
our purpose perfectly.
Thanks,
Oleg
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
