[xmlvm-users] Best Practice and Incremental Conversions

[<ju...@ed...>]

I am seeking some advice on how best to prepare our Android -> iOS  
conversion(s).
We want to minimize effort and risk.

I would like to create a **known good conversion** of each of the  
binary artifacts in our real-world situation.
Here is a good description of our binaries:

DESCRIPTION			NOTES					ARTIFACT
===========			=====					========
External Library : 							xstream-1.3.1-POS.jar
OurOwnCustomClasses_1 : 	code that uses xstream -> 		Custom_1.jar
OurOwnCustomClasses_2 : 	other code -> 				Custom_2.jar
OurOwnAPKLib :			Custom_1.jar + Custom_2.jar		Custom.apklib
OurOwnAPK :			Custom.apklib + Android + AndroidDeriv	Custom.apk

After trying a few times with the big-bang approach to "convert  
'Custom.apk' to 'Custom.iOS'", I realized that to make this actually  
work, i need to deal with each binary separately.
In the big bang approach I got errors like:
<quote>
-xproject-create:
    [delete] Deleting directory /POC-1/build/.xcode
      [echo] Creating Xcode project with the C backend.
      [java] [09/17/12 10:29:09.620]    ERROR: Couldn't create node  
for com.thoughtworks.xstream.XStream
      [java] [09/17/12 10:29:09.622]    ERROR: Couldn't create node  
for com.thoughtworks.xstream.XStream
      [java] [09/17/12 10:29:17.120]    ERROR: Vtable for  
com.thoughtworks.xstream.XStream not found!
      [java] [09/17/12 10:29:17.120]    ERROR: Vtable for  
com.thoughtworks.xstream.XStream not found!
</quote>

Now, I immediately started to think about breaking the work down bc  
XStream is a rather complicated library.

So, I attempted 2 conversions:
[1] xmlvm --in=xstream-1.3.1-POS.jar --target=xmlvm --out=xmlvm-output/xmlvm
[2] xmlvm --in=xstream-1.3.1-POS.jar --target=iphone --out=xmlvm-output/iphone

And, I got no errors in each case ! Nice !! Here is a summary of the  
generated contents:

xmlvm-output/xmlvm :
from 342 class files in the xstream jar we ended up with 343 .xmlvm files

xmlvm-output/iphone/build/xcode/src/app:
from 342 class files in the xstream jar we ended up with 684 files  
(342 .h and 342 .m)

So, the outputs look promising.
Of course, I do not not which artifacts are the ones I should consider  
using :)

So, my thought is to perform incremental xmlvm (and/or iphone) build  
working down the ARTIFACT list above.
This would mean that xmlvm executions would successively have an  
increasingly larger number of 'pre-compiled' inputs.
It also means I know where errors/issues are at the library level.
(We would probably manage the 'successful pre-compiled derivates' with  
Maven - as we do our Java outputs and dependencies.)

So, if you have experience of such an approach (not on the Maven side,  
we are ok with that), could you please advise me on how to structure  
my builds.
I am guessing that care would need to be taken so as to avoid  
duplicates, and as I do not understand the internals of  
dependency/input management in xmlvm, I would need advice specifically  
on this and how to create meninful successive invocations of xmlvm.

Best Regards
Julian