Re: [ajdt-dev] Preparing for AJDT 1.1.3 release - status

["Adrian Colyer" <adrian_colyer@xxxxxxxxxx>]

I've fixed bugs 37033, 37037, 37762. Following some investigation from Julie, I was also able to commit a fix for the missing class files after an incremental compile. In doing this I uncovered a serious flaw in AJDTs support for incremental which is now fixed - hopefully this will make incremental a much better experience.

<detail>
We were missing a call to IResource.refreshLocal(..) [that's the api you were hunting for Julie] on the project directory after a compile. This was compounded by the more serious problem that we were always reporting that in incremental compile had finished (rather than waiting for it to actually complete) - thus there was a race between the compile finishing and the post-compilation actions taking place. With this fixed, incremental mode feels more stable and predictable again.
</detail>

That leaves me with 37035 and 37948 to go. 37948 appears to be another instance of the SWT error / deactivated plugin problem that you were looking at Julie (there's a stack trace in the bugzilla report). It will be tomorrow before I get to look at that, but in the meantime my suspicion is that a static initializer somewhere is perhaps making an SWT call? Interesting that the trace kind of suggests that its actually JDT rather than AJDT that blows up, but we have some more work to do yet before we can claim our innocence!

I've validated that an erroneous line in a .lst file does indeed now correctly report an error. Would be even better if the error could be reported against the .lst file (for double-click navigation) rather than against the project. I suspect this would need an Ajde fix and so will have to wait for  AspectJ 1.1.1 right Mik?

-- Adrian.
adrian_colyer@xxxxxxxxxx


09 June 2003 10:50
To: ajdt-dev@xxxxxxxxxxxxxxx
cc:
From: Adrian Colyer/UK/IBM@IBMGB
Subject: [ajdt-dev] Preparing for AJDT 1.1.3 release

Notes from the 
developer call last Thu... 

Mik to 
work on the following three items as raised in his post to aspectj/ajdt 
-dev: 

- Crosscutting structure model 
incompleteness: inter-type declarations info
is missing, advice on 
constructions is missing. 
- Parsing of "../" 
paths in ".lst" files is broken.
- There is a new suspicious bug causing 
*all* Eclipse compiler warnings to
show when no "-Xlint" option is 
passed.

From Andy's post to 
ajdt-dev, Adrian will look into points 1 & 3: 

1) LST files with bad entries cause compilation to 
terminate prematurely. (validate this is now fixed) 

3) In incremental mode, sometimes our builder just stops 
getting called. 
The builder keeps a log 
of how it is used in the AJDT event trace.  I find 
after a while it just stops getting AUTOBUILD calls for 
some reason. 

Julie to investigate 
#2:   

2) Investigate James 
Howe's report about the workspace becoming 'corrupt' 
for no apparent reason - I've seen this on 1.1.1 and 
appear to have 
alleviated it by starting 
eclipse with more memory - but thats not the 
ideal solution. 

Julie to 
investigate Chad Wooley's problem with classfile creation. 

All bugzilla bugs scheduled for fixing this release are 
assigned to one or other of the team (see bugzilla). 

Final task will be to update the docs from the 1.1.0 
aspectj release. 

-- 
Adrian
Adrian_Colyer@xxxxxxxxxx