There has been talk recently about a ReactOS newsletter in the tradition of Kernel Traffic or WineHQ. At the same time I began investigating the ReactOS scene in hopes of helping. Unfortunately, my C/C++ skills are not up to par. Steven Edwards suggested I work on the ReactOS newsletter, which until this point, was absent from the ReactOS scene. Thus Splash was born.
Splash is the official ReactOS newsletter. Each issue will highlight and cover the interesting points that happen on the mailing list. It will also cover, when I can decipher it, the activities on CVS and IRC. Issues are planned to be released on a roughly bi-weekly basis. This project will be a learning project, a get-better-as-time-goes project. I have little experience with writing technical newsletters, and even less with ReactOS itself. Fortunately, time should help my judgment and writting. I hope you enjoy Splash. If you have any questions, comments, concerns or hate mail, please send it to: reactos (at) tong-web (dot) com.
An alternate layout of Splash may be found here
David reports that he has been unable to install ReactOS with DBG=1 and KDBG=1 enabled. This happens because KDBG is designed to treat all exceptions as kmode exceptions. In this case, the VMware detection routine attempts to perform an instruction, that on normal systems, would raise a umode exception (or nothing at all if on a VMware system). Unfortunately, KDBG treats the umode exception as kmode, and goes *boom*, regardless of it is actually running on VMware or not.
At the time of this writing, Art said he would commit a change to KDBG that would allow configuration of umode exception catching. General consensus (on IRC) seems that this is a good idea, and furthermore, configurable at runtime rather than compile time.
Gerard conducted some testing on the Realtek 8139 NIC. He reports that the detection/enumeration of the third PCI bus (PCI BUS 2)
has been fixed (thanks to Eric Khol, bug
#436). Additionaly, he reports that the RTL8139.SYS loads successfully
but does not start properly. This is due to a unimplemented function in ROS, "NdisMQueyAdapterResources" (new bug,
#446). Filip offered a workaround (
io_res_patch,
binaries) as well as an
"example filter driver that implements PnP Arbiter for Windows"
(available
here). Filip also requests that someone adapt and implement the code for
ReactOS, as he doesn't have time.
Despite the workaround, it appears that the driver still does not start properly.
The ne2000.sys bug that was previously causing problems has now been resolved by Arty. A structure in ne2000.sys, named PACKET_HEADER, was previously thought to hold the ethernet frame header. After some digging, Art discovered that it is merely 4 bytes and only held which buffer page to use next. This was assigned by the ne2000 hardware, and had nothing to do with the ethernet frame itself.
Art says he has
"now factored in the ethernet frame header and made tcpip work the right way, that is to remove consideration
of the ethernet frame header at layers above lan in the receive pipe. This should make all adapters work the same way for larger
packets (and ne2000 now work right)."
Art used what Filip said, as well as examples from here, to fix the bug.
Gé van Geldorp noticed that Ekush( Google Cache), a project designed to do the same as ReactOS, had published its binaries. After some simple string searching in the binaries, it was determined that Ekush was in GPL violation for using ReactOS source without abiding by its license.
Jason Filby said he would write
"a letter explaining [ReactOS's] position (when their site is back up) with appropriate links to fsf.org and
perhaps eff.org."
Thomas Weidenmueller describes why Ekush is violating the GPL, saying,
"Remember, we still own the copyright, we just permit others to use the sources as the license
says. Which, as I understand it, implies they must not remove copyrights, neither from binaries nor from sources."
After some
more investigation, Filip Navara discovered that they had also included
QEMU(licensed under LGPL). QEMU was renamed to EmuPC. There were references to
FreeType and Wine as well.
But the plot thickens. Gé van Geldorp reports that shortly after the first email was sent to the ReactOS mailing list, the site went "Down for Maintenance". The next day, the site came back up with new binaries. The references to ReactOS mentioned by Gé van Geldorp were all gone, minus a few. Geldorp submitted a story to Slashdot which highlights the code theft (available here) and Anich Gregor put together a list of identical ASM code between the ReactOS and Ekush (which can be found here). It appears that the Ekush website is now down. Whether done by Ekush itself or its hosting company is not known.
This section contains nothing more than interesting CVS comments that I personally like. There are more CVS comments then listed here, and many more actual commits.