Community
Participate
Working Groups
This is becoming a big problem on a current project. Quite often (more so after switching to 1.5.1+, it seems), a developer will call me with a problem that an aspect should have taken care of, but clearly isn't. After looking at code, it seems all fine. So in a dismay, I fire a clean compile and the problems goes away. I wanted to see if there is a pattern here, but failed to observe any relationship between the kind of changes leading to the problem. The general characteristics only tell that the classes, and not the aspects are changed before we observe this problem.
The most common problem seen is an advice that stops being applied to a join point (method execution) after an incremental compilation, but works fine after a clean build.
I do know of a change vaguely in that area during the 1.5.1 timeframe. Can you possibly turn on all the debug that you can for the AJDT event trace view? When you make a change to a class file it should report some information about the decision process it is following - why certain things are being compiled and any knock-on effects of that. Using your knowledge of what the change was, does the list of changes make sense?
Ramnivas - is this still happening for you? Did you get a chance to try creating the AJDT event trace that might indicate what is happening? And its *definetly* that the code on the disk is not woven, yes? Not just that the markers are missing?
marking up the severity so I can see it in the big list-o-bugs.
It hasn't happended in a while (to make matter worse, it used to happen on other developer's machine, and I have less control over its environment). I am still keeping my eyes on this problem. When it happened, however, the problem was advice not getting applied in byte code (and not just missing markers). Basically, the functionality implemented with aspects was absent, leading to bizzare behavior (and often crash). On every occasion, full rebuild solved the problem.
any sign of this bug recurring ???
Nope. Also, the project has gone into functional QA testing so not much is being changed. That poses the problem of what to do with this bug. May be mark it as "later" and reduce is severity (due to non reoccurance)?
I'm ok with that - if you see anything like it again we can resurrect this bug.
LATER/REMIND bugs are being automatically reopened as P5 because the LATER and REMIND resolutions are deprecated.