вівторок, 27 липня 2010 р.

JSR 45 : WTF, where my smap info ?

Design of JSR-45 supports usage of source-information in generated java bytecode, but not in generated sources. When most of java frameworks are over-designed, this is under-designed. Trivial case: producing of java source which is processed by standard javac, does not covered at all. This cause not only need of extra tool to install SMAP, but more painful issues: errors in generated java file does not handled by javac correctly. Therefore if you using tools, which generate plain java files (like javacc or our jpe), you still need look at generated source codes. Possible workaround: create a tool, which run javac, then substitute source files and line-numbers in javac error message using preprocessor-generated smap files.

вівторок, 12 січня 2010 р.

Temporal relations and O/R mapping

Just finished first small project, which use ri-tree at the same time with traditional ORM mapping.
Platform result: hibernate plugin .

NMain problem was absence of support for parametrized domains in Ejb/QL from clause.

вівторок, 7 квітня 2009 р.

Syntax patterns for micro-optimization.

More about project coin: effect from some (not all) proposals can measures by number of LOC-s, which can be rewritten using new technique.

Some results and notes in coin mail-list:
(text)

Source code of patterns:
(browsable svn)

середу, 25 березня 2009 р.

It's my project coin proposal for enhanced string literals in Java . (Just to keep record about closed project).

On reject of this proposal we can see difference between corporate and true open source: I. e. near any open-source changed adopt such extensions or, in case of differences in points of view: clear state why this is not acceptable (and propose way to integrate similar or alternative ideas in future).

In case of corporate OSS, we does not see, not adoption, not clear statement about using alternative way. It's why bazaar such effective, and trying to simulate open-source without
bazaar actually does not give us bazaar-like speed of development. (More pity - such simulation
also generate frustration of potential contributors: If I donate my time, I expect that my effor will be or used in some additive way, or rejected on early stage [before any serious time investement], but not thrown out without any result) It's why language stagnating last few years.

About need in enhanced strings: look at references section. I. e. we have
- 3 proposals, raised independently (I discovered this after issuing own proposal)
- 2 'hacks' for implementations

вівторок, 17 березня 2009 р.

OpenJDK build

It's hove one differ from documentation:

Problem #0

forest extension to mercurial
Actually exists two forks
a) http://bitbucket.org/pmezard/hgforest-crew
(work with python-2.4, does not work with python-2.5)
b) http://hg.akoha.org/hgforest
(work with python-2.5, does not work with python-2.4)

Yet one pity news - default htwebdir.cgi does not support forest.


Problem #1:

1. builds breaks, as described in

http://forums.java.net/jive/thread.jspa?messageID=234284

(quick and dirty workaround was copy rt.jar from JDK-1.6 to /jre/lib/rt-closed.jar)

Still compiles.

вівторок, 10 березня 2009 р.

Sun OpenJDK mercurial is down ?

Trying to clone openjdk repository to my local copy:

hg fclone http://hg.openjdk.java.net/jdk7/t1 t1
abort: requirement '<body bgcolor="#f0f0f8"><font color="#f0f0f8" size="-5"> -->' not supported!

Yet one problem - openjdk web pages does not list any contacts, where I can fire about not working repository.

UPD: and the problem was on my side: root directory to jdk7 forest is not t1 (t-'one'), but tl. ('t-l')

четвер, 7 лютого 2008 р.