wisberg [Sat, 24 May 2003 09:16:42 +0000 (09:16 +0000)]
updates mainly to support maven but also to support new ICommand wrapper for AjcTask:
- findToolsJar now accepts versioned jars
- supporting a forkclasspath to permit user to explicitly set
- upgrades for testing
- so multiple iterations work,
- not altering cmd in execute()
- only fixing up outjar once
- reset() complete
- so incremental compile tests work,
- passing quit through to main
- when forking on windows, use javaw rather than java
if possible to avoid popping up a console window
wisberg [Wed, 14 May 2003 05:35:38 +0000 (05:35 +0000)]
defensively supplying ISourceLocation.NO_FILE when there is no file in the result.
This is needed for file-based message matching to not fail when no file is specified,
and might provide a more sensible error message if the file is used.
This only replaces a NPE or invalid (empty) filename strings.
wisberg [Wed, 14 May 2003 05:17:49 +0000 (05:17 +0000)]
- now supports both eclipse bin dirs (default)
and ant-built test jars
- added incomplete test-all-at-once
(currently runs each from its module directory)
wisberg [Wed, 14 May 2003 05:08:45 +0000 (05:08 +0000)]
- new data (IMessageHolder rather than lists), same algorithms
- using medium-strenght message comparison that includes soft
comparison of file and text
- unexpected info messages not marked as error
- expected FAIL+ messages suppressed to avoid false negative
wisberg [Wed, 14 May 2003 05:06:47 +0000 (05:06 +0000)]
- enforcing -incremental => staging
- permitting bad input using badInput="true"
(though still checking during staging)
- permitting -XserializableAspects
- supporting new AjcMessageHandler interfaces
- supporting parent badInput in XML
wisberg [Wed, 14 May 2003 03:13:28 +0000 (03:13 +0000)]
No functionality change.
This just makes explicit that the browser reads user preferences
rather than relying on the default value supplied by the default constructor.
wisberg [Fri, 9 May 2003 22:52:28 +0000 (22:52 +0000)]
Another strange incremental test, taken from ajcHarnessTests
and comparable with a known limitation case for injars,
this sometimes doesn't fail for me. hmm.
(requires pending changes to sources)
jhugunin [Fri, 9 May 2003 17:31:49 +0000 (17:31 +0000)]
fix to use cache for isPackage method (as indicated by previous !!! comment)
this fixes the two failing incremental test cases just added by Wes and now
moved to ajcTests as passing
wisberg [Fri, 9 May 2003 07:24:29 +0000 (07:24 +0000)]
Backing out the change I made to add the output directory to the classpath.
This is wrong because we should resolve types wrt the cached classes in the weaver.
But if this change were left in, the test cases just added to ajcTestsFailing.xml pass.
wisberg [Fri, 9 May 2003 07:14:19 +0000 (07:14 +0000)]
Fixing Harness bug of adding the output directory to the classpath, which conceals a bug in the incremental compilation.
This change should have no effect on other tests, but I noticed two other failures during normal compiles caused by inability to resolve type names.
wisberg [Thu, 8 May 2003 08:39:31 +0000 (08:39 +0000)]
adding output dir/jar by default to classpath to permit resolving types in incremental compile
Might need to be more discriminating and/or to document.
wisberg [Thu, 8 May 2003 08:31:21 +0000 (08:31 +0000)]
- removed invalid test requiring -sourceroots in the argument list. They might be in a config file.
- handling request for fresh rebuild during incremental
wisberg [Thu, 8 May 2003 08:05:01 +0000 (08:05 +0000)]
updated with tests, etc. Not working yet:
- junitModules.xml needs better setup with testable jars
- cvs unvalidated
- ajcTests unvalidated - see also ../build.xml script
wisberg [Thu, 8 May 2003 08:03:30 +0000 (08:03 +0000)]
incomplete update to support module jars rather than eclipse bin directories when running junit tests
the old way with bin directories still works fine.
wisberg [Wed, 7 May 2003 10:07:20 +0000 (10:07 +0000)]
supporting a forked java run provisionally with system properties specifying fork, java home, java, and bootclasspath.
This should enable running the harness when targeting a VM that the harness or compiler can't run on,
but this does not detect failures correctly in some cases,
nor does it validate any of the system properties.
wisberg [Wed, 7 May 2003 04:02:16 +0000 (04:02 +0000)]
- support for "same" tag, which makes no changes before incrementally compiling
- support attribute for a fresh rebuild, currently implemented only by ...ajde.CompileCommand