trouble building mlvm

[Paul Phillips]

On Tue, Dec 16, 2008 at 12:09:27PM +0100, Iulian Dragos wrote:
> I'm trying to build a version of JDK7 with invoke dynamic support.

Hey Iulian.  Me too.  Unfortunately I took on the added excitement of trying to do it on OS X, 
where the main bsd-ports tree is broken, so I have to apply an additional patchset from 
Michael Franz to fix the int32/intptr issues.

I had pretty much the same experience as you, but I think I got somewhat further.  After 
seeing that the web site instructions didn't actually apply the patches, I applied them in 
order by hand, i.e. by calling patch myself.  Most hunks were successful; most failures were 
easy to track down; a couple required some head scratching.  I eventually got them applied, 
but the build failed late in a way that made me suspect I'd made a bad decision, most likely 
in the callcc patches.

With my much improved knowledge of the whole process, I was going to undertake it again from 
scratch like so: check out the jdk sources, and revert to the last revision the mlvm patches 
are checked against (which should be b38) in the jdk, hotspot, and langtools repositories:

  hg revert --all -r jdk7-b38

Then apply the mlvm patches, and finally use hg to merge forward to b40 or b41, wherever we 
are now.  Unfortunately I hurt my neck yesterday and I think this email is about all the 
computing I'm going to be able to manage today.  I too am a total mercurial novice which makes 
this much harder; I'm gitifying the forest next time first, so if I do get this working (and 
can keep the osx patches distinct from the mlvm patches) I'll be happy to send you patches.

P.S. Thanks for fixing "return from lazy val".

-- 
Paul Phillips      | A Sunday school is a prison in which children do
Imperfectionist    | penance for the evil conscience of their parents. 
Empiricist         |     -- H. L. Mencken
i pull his palp!   |----------* http://www.improving.org/paulp/ *----------