aclement [Wed, 15 Feb 2006 09:41:19 +0000 (09:41 +0000)]
optimization: made unpacking of the table lazy. With a minor change in AJ itself, this results in us only unpacking 15000 of the 39000 local variable table objects constructed during a build of shadows. Saving space and time.
aclement [Wed, 15 Feb 2006 09:15:34 +0000 (09:15 +0000)]
optimization: made unpacking of the table lazy. With a minor change in AJ itself, this results in us only unpacking 15000 of the 39000 local variable table objects constructed during a build of shadows. Saving space and time.
aclement [Tue, 14 Feb 2006 13:32:07 +0000 (13:32 +0000)]
optimization: rather than continuously constructing new arrays from the linked lists of instructions, we create one array and use it multiple times in MethodGen.
aclement [Mon, 13 Feb 2006 15:37:17 +0000 (15:37 +0000)]
optimization: use bit twiddling rather than Sets for manipulating the list of Pointcut.couldMatchKinds - (saves a few percent when building shadows in my tests...)
aclement [Tue, 7 Feb 2006 15:17:06 +0000 (15:17 +0000)]
lazy unpacking of the line number table contents. Only done when required - usually when a MethodGen built for a Method (i.e. we are going to weave it).
acolyer [Fri, 3 Feb 2006 11:52:01 +0000 (11:52 +0000)]
Progress on: Bug 126328: IlegalStateException : zip file closed
https://bugs.eclipse.org/bugs/show_bug.cgi?id=126328
closes the loop hole with >1000 jars, and adds some extra sanity checks on potentially closed zip files (tested that the call is cheap) to give better error messages.
acolyer [Fri, 3 Feb 2006 10:16:11 +0000 (10:16 +0000)]
Progress on: Bug 126316: Invalid method signature: Pjava/lang/Enum<Ljava/lang/Object;>;
https://bugs.eclipse.org/bugs/show_bug.cgi?id=126316
--> test case to reproduce
acolyer [Tue, 31 Jan 2006 18:13:24 +0000 (18:13 +0000)]
made all the state in AjState private so we can get a better handle on what's going on. Added accessors where needed and extensively commented the lifecycle of the key state we hold.
aclement [Thu, 26 Jan 2006 10:57:12 +0000 (10:57 +0000)]
necessary magic. Factory 'mechanism' used so that the JDT project doesnt have a direct dependency on the ajdt.core project. It references the factory name through a string and dynamically loads the type.
aclement [Thu, 26 Jan 2006 10:55:44 +0000 (10:55 +0000)]
Subclass for the AST that is able to create pointcuts/declares/etc. Similar ctors to the supertype and one method copied into this class that was creating the AST directly when it needs to use the factory now to ensure it creates an instance of this subtype.
aclement [Thu, 26 Jan 2006 10:54:40 +0000 (10:54 +0000)]
new placeholder nodes for type patterns and signature patterns. Before this change this info was being 'discarded' and not included in the constructed AST - now it is at least created and held onto. Future work would create the various subnodes for the different kinds of typepatterns and signaturepatterns - if users want to start analysing them below their 'top level'
acolyer [Fri, 20 Jan 2006 12:28:01 +0000 (12:28 +0000)]
ant script for running profiling tests against aspectj with ajc source compile, binary weaving, load-time weaving, and AJDT simulation. Can configure target application and source of compiler (workspace or pre-built jars).
acolyer [Wed, 18 Jan 2006 14:37:02 +0000 (14:37 +0000)]
an additional interface that custom message handlers may choose to implement. Implementors get callbacks when a build is starting and when a build has finished (allowing them to perform before and after build processing, like closing output files etc..).
acolyer [Wed, 18 Jan 2006 14:35:44 +0000 (14:35 +0000)]
Use ant's TaskLogger for all output from our task, rather than writing directly to sys out / err. This commit also fixes the issue that running ant -verbose did not produce verbose output from iajc (it does now). I ran into this with a client yesterday, and it had me stumped for a while!