it appeared that the kernel's sysGlocals() structure begin to be badly allocated, leading to random bugs like i386 debug handler fault, page fault handler corruption, etc. This was due to the fact that the sysGlocals and kGlocals structure for the kernel initializor thread were still allocated from an old ASM file that didn't knew the struct has grown up that much ...
Now, the structure has been moved to src/kernel/init.c and therefore its size will reflect what is described in "glocals.h".... read more
ENOUGH STATIC VARIABLES! There is kdsInterner, kdsVsetClass ... Now all these will be encapsulated in a preferences (profile) of the KDS environment: the KDS domain. kdsDomains should also allow the creation of sub-space in KDS that will behave differently (for instance allocating objects in process-specific area rather than microkernel area) and will further help defining cross-domain calls to KDS, by mounting/unmounting domains and relaying the lookup/register/unregister/rlookup calls to the correct domain.... read more
most of the source code for events processing is written, tests are imminent...
hi there ... i've been so busy that development has severely slowed down :-( I wish a happy new year 2003 to all of you.
I'm currently working on the implementation of Events Processing Engine for Clicker microkernel. This is a mandatory step before we can implement drivers and will open the path to true IPC communication without requiring excessive resources from the microkernel area...
it seems that capitilized filenames doesn't please that tool that complains "makefile" is in the way when updating "Makefile" ...
after 2 monthes of hard work in the CVS (cvs.clicker.sourceforge.net), we finally have a complete source tree that allows compilation under both Linux, Dos & Windows, including floppy-disk creation tool for Linux users ... enjoy it at http://prdownloads.sourceforge.net/clicker/c32-lxsdk-021117a.tar.gz?download
The SDK is going on ... now the kernel has been built successfully from the CVS repository both under Linux and Windows 9x. We're still missing a make floppy command, though :(
the CVS repository becomes to look like a big mess .. i've been trying to upload the SDK from home (simply updating the local repository and then committing), but it doesn't work ... i'm afraid there won't be CVS update before i get Internet@home back :(
With the last debugging session (woo, took a long time before i moved my CPU to my new flat ;), the ThreadKill Bug of XLR8 extension (a.k.a. semdemo.unload() bug) is fixed. Still a few regressive testing is needed and then i'll make a full (binary/floppy/source) release of the XLR8ed kernel ... stay tuned :)
Hi dudes, i'm moved in my new flat, but i haven't moved my computer yet, so expect me to be slower to respond to e-mail sent past 16pm GMT
this week, we welcome a new member, Stain! He started working on Vesa & mouse support already, and has also a personnal OS project on the workbench ;)
hope we'll learn a lot from each others.
an alpha version the DjGpp SDK is now available. it should be more easy to install and configure for your system. Just ensure DJENV variable is set, and that CLICKER environment variable reflect the SDK installation directory ...
Hope it will help incoming members to get things compiling more quickly...
Sorry if you're trying to mail me: i'll be at > 10kms of any kind of computer or Internet connection for the week ...
it seems that it is too complicated to install Clicker from the sources in DOS environment. An "all-in-one" distribution could be helpful, but it would also be fine if we could develop completely (from source editing to floppy writing) under the penguin-system ...
Unfortunately, this requires some tools (such as progmaker, modmaker, coff2sfs) to be re-written for the ELF file format ...
Well, dudes, my "'In Real Life" world is taking over my geek world :) more heart beats and fewer CPU cycles .. I'll be back in a few weeks, but i probably will have less time to give to Clicker, so i'll rely on every other members to make the project go on ...
Come on, dudez: i know you _want_ the project to move on: it's your responsibility too : check the Web for some useful docs, try out hardware combinations, etc. You have a powerful microkernel and a cute interface to develop modules for it, so what are you waiting for ?... read more
Well, that was a long time since the last complete release. A lot of work has been required to move to the XLR8 architecture as it touched to a very hard-to-debug set of features like task switching, stacks manipulations, dispatching, etc. However, the resulting code is probably a lot faster.
In addition, many modules are now converted to KDS2, and thus have a more straightforward (and less dangerous to use :) interface to kernel or other modules, thanks to the automatic .api generator (which only handles automatic imports/exports for now).... read more
The conversion of tests module have been started. Cleaner, smaller, they take advantage of the new automatically-generated .api files holding conversion rules between kernel symbol names (kprint) into kds-cpp conventionnal names that modmaker2 will be able to interprete and translate.
Finally, we could keep pure C even for KDS2 automatic symbols (that were initially suspected to require C++): a new Perl script can sweep exports.c file and create a translator .api file for any module.
Perl will catch the KDS path and automatically generate the same weird name as C++ would have done, and create a
#define <kernel symbol name> <KDS-C++-encoded name>
So, your module will just have to #include <api/kernel.api>, and then every KDS-exported symbol can be used exactly as if you were programming directly in the kernel.... read more
The XLR8 extension for Clicker multithreading support is now almost ready. Some modules are still to be upgraded to support the new architecture but the microkernel can already handle both stack-switching (faster) and task-switching (more secure, hardware-support for address space/protection context switching.)
Well, activity is going downward these days: heavy pre-paper submission stress at work (for IWAN 2002), for me, and most probably exams time for most contributors :)
Moreover, the 'xlr8' thread architecture is tricky to debug ... stay tuned
the nasty bug #559037 has finally been terminated B-)
watch a CVS update within the day.
During the same time Rajeev has been busy writing a teacup on PCI framework - actually under review - that will be put online at http://clicker.sourceforge.net/docs/cake/pci.html
Huge thanks are flying to the Pro-POS team (especially Solar and Devl) for their CVS documents and SDK: thanks to them - even if they don't know it - we now have a CVS for clicker :)
Trumpets! Drums !
Happy people, i proudly announce you that Clicker's next release (and the actual CVS revision) won't rely on a crappy ASM code to perform deferred task switches or processing events. Result is a cleaner code that actually seems to achieve better responsiveness.
First documents produced with the Doclang are available at
http://clicker.sourceforge.net/nfo/einstein/einstein-module.html
and
http://clicker.sourceforge.net/nfo/kds/kds-module.html
the compact documentation that has been used to produce them is uder nfo/einstein/einstein.nfo and nfo/kds/kds.nfo respectively :)
rajeev will work on the implementation of the Events processing engine. Welcome to him !