Tuesday, July 22, 2008

[Radeon Update] Catalyst 8.7 saves the day!

Ok, the Catalyst 8.7 are out there from a few days and I installed them today. Big surprise! After enabling again DRI the system doesn't freeze! Wow: I'm impressed to be out of the darkness. Let's resume the situation:

+1:
  • Installation works out of the box (have to remember to rerun it after each kernel update ;))
  • Accelerated X works
  • Compiz works (get you transparent windows and video effects)
-1:
  • Terrible flickering during windowed video playback (switching to fullscreen gives HW-accelerated, flicker-free video playback though). Workaround exists but it makes full screen quality very bad (large pixels anybody?).
  • The small preview thumbnails do not show. That is, when switching with ALT+Tab do, there are small white squares instead of containing the window' thumbnail.
About the performances I could not say:
mario@gear:~$ fgl_glxgears
Using GLX_SGIX_pbuffer
18093 frames in 5.0 seconds = 3618.600 FPS
19450 frames in 5.0 seconds = 3890.000 FPS
19352 frames in 5.0 seconds = 3870.400 FPS
19265 frames in 5.0 seconds = 3853.000 FPS
19168 frames in 5.0 seconds = 3833.600 FPS
Anyway, it is enough for a satisfying Compiz experience!

I hope that the minor issues are going to be fixed soon. They are minor glitches but they ruin the ability to fully enjoy this card under the penguin.

Saturday, July 19, 2008

Catalyst 8.6 + Linux + 8Gigs of RAM = Crash

To introduce just another chapter in my neverending fight to have a working system, I'm going to tell you about how I've been able to make by Ubuntu Hardy (8.04) box start (yeah: no freeze!) along with my new Radeon HD 4850.

I previously owned an nvidia GeForce 7800 GTX 256M an decided to change it to enjoy more recent games: at 135,00 € the Radeon was too big an attraction. The bad things did happen when I tried to start X after following this guide to install the latest 8.6 drivers (supporting the new Radeon 48XX series cards). I followed step 2, rebuild the .deb packages and installed them. Restart X and ... freeze! Hard reset needed. From there on several ours went on googling, searching forums and reading bug reports and trying any possible solution:
  • disabling the SMP with maxcpus=1 (maybe some concurrency lock? Difficult since there the crash occurs deterministically bu you never knows ...)
  • reducing the visible amout with mem=2048m (maybe some overlapping among memory and I/O addressing?)
  • disabling 3D acceleration (also if I'm pretty sure it isn't used by GDM but maybe some strange initialization hard locks the machine ...)
Nothing.

Though, I'm pretty sure that it is a combination of 64 bits CPU, 8G of RAM and the ATI drivers. Before giving up, I came over a bug report: essentially the users complained about the same hard freezes as me and the only way they could boot was by disabling DRI altogether, that is by adding:
Option "no_dri" "yes"
in your /etc/X11/xorg.conf file(video card device section).

The good thing is that you finally have a working X. The bad thing is that there is no HW acceleration and dragging windows is sluggish, at best. Hope that ATI is going to fix this because I can live without OpenGL under Linux but dragging windows around is a so common that such behavior can't be accepted. Perhaps I should switch back to NVidia, since their prices are going down.

Thursday, June 26, 2008

Playing in the sandbox

If you want to run your code in a secure environment (a sandbox) within control of the JVM Security Manager you can take a look at here. This is particularly interesting for me that I'm quite unaware of the API. For more links about security within the Java Platform, look at this page.

Sunday, June 22, 2008

Tracking the crash


Did you remember the bug about the the JVM crashing after running eclipse for a while? The problem seems to be due to a bug well into the JDK. Infact, I've installed Eclipse 3.4 RC4 on my brand new Windows Vista x64 installation and ... bang! Got it! So it is a really multi-platform bug for 64 bit platforms. And I can spare you a google search: it is not fixed yet (JDK 1.6.0_u6) but you may be willing to keep your eyes on it at SUN bugtracker.

Sunday, June 15, 2008

Penelope is born!


Penelope is an Aspect-based framework for enhancing Eclipse RCP applications. It leverages several 3rd party products like Equinox Aspects and Guice/Peaberry and aims to provide a way for creating component-based RCP applications using test-driven-approach.

But more importantly, I've finally started to finalize all this Aspect-related stuff in something really useful! Next moves will involve:
  • verify the feasability of having a platform-wide injector instead of several for bundle.
  • infrastructure for Model-View-Presenter-based development for Eclipse RCP: together with the Presenter-first approach we want to support test-driven development with no excuses! This will require extending the Guice/Peaberry framework with RCP-related support classes.
  • support for Eclipse Extensions (possibly in collaboration with Peaberry)
  • some kind toolbox integrated within the Eclipse workbench to ease development tasks.

Thursday, June 12, 2008

[Solved] Eclipse 3.4 + Ubuntu 8.04 + SUN JDK == Crash

I've been trying to run eclipse 3.4 RC3 on my ubuntu 8.04 64 bit box: nothing. The issues can be tracked to two things:
  1. XULRunner should be updated to latest version, otherwise the internal browser will crash, for example when showing the Welcome view.
  2. The workbench will crash when attempting compilation of your workspace for the first time. This seems to be related to a bug in the HotSpot of your JRE: if you take a look at the crashlog and see something like:
Current CompileTask:
C2:1504 org.eclipse.core.internal.dtree.DataTreeNode.forwardDeltaWith([Lorg/eclipse/core/internal/dtree/AbstractDataTreeNode;[Lorg/eclipse/core/internal/dtree/AbstractDataTreeNode;Lorg/eclipse/core/internal/dtree/IComparator;)[Lorg/eclipse/core/internal/dtree/AbstractDataTreeNode; (469 bytes)
then, welcome into the club ;)

Fortunately there is a workaround even for this by adding

-XX:CompileCommand=exclude,org/eclipse/core/internal/dtree/DataTreeNode,forwardDeltaWith

in your eclipse.ini JVM options.
Kudos to the great community! :D

References

Wednesday, May 14, 2008

E4

Eclipse 4 (codenamed "E4") is being discussed by the architectural board on May 15. You will note that Dependency Injection within Eclipse is also one of focal points. Yeah! Just now that we resorted to use Peaberry + AOP in eConference ;)