Bugzilla Bug 145568

It tests presence of the following on classpath:
jimi (sun-jimi), jai (sun-jai-bin), trax (xalan), jdk14, jce

This is bad because jimi and jai are controlled by useflags (linking
appropriate jars into lib/ dir based on the flag), but the autodetection can
mess with it because ant can be using jai itself, and that can result in (with
USE="-jai" for fop):
- java-strict is on, but the autodetection will still see jai on classpath when
doing the test (-Dbuild.sysclasspath=ignore doesn't affect this, sadly), during
compiling jai will be hidden and build will fail
- java-strict is off, it will build jai support because it's in ant's
classpath, but it won't be recorded as dep in fop's package.env

One solution can be to put ignoresystemclasspath="true" parameter in
<available> tags in build.xml used to test jai and jimi presence. This will
ensure that jai and jimi support will be checked only in lib/ dir, not on
classpath ant uses. This could be done even without patching, by running
xml-rewrite.py from ebuild. But since there's already a build.xml patch, I
expanded it to do the following:

Instead of testing for jdk14 and jce, it enables is unconditionally. Also
changed the dep from || (jdk1.4 jdk-1.3) to just jdk-1.4.
Same with the trax testing, it's just enabled, because xalan is unconditional
dep anyway.
Tests for jai and jimi are removed and the associated properties are not set in
build.xml at all. Ebuild passes these properties based on useflags.
No automagical detection is left :)

Created an attachment (id=95444) [details]
new patch for build.xml

Created an attachment (id=95447) [details]
patch for fop's ebuild

Released in fop-0.20.5-r7