Activity summary

[b4dc0d3r]

Experimental summary of SVN activity. I may keep updating/editing, I may not. If you don't understand something, it's probably not of much interest to you, or you can use the intartubes to find more info. Some copied verbatim from SVN log, mostly not. I may have some of this wrong, and future commits might undo some of this, but it shows what's happening right now.

Commit numbers 46622 - 46658, trunk unless otherwise specified.

Kernel security - fix locking in access checks, work around broken objects, better checks

HAL - ACPI enhancements including Plug and Play (requires fixes in PnP), split into PC/AT and ACPI versions and some accounting for differences between the different versions

Ks (Kernel Streaming) - Improvements for devices, filters and pins. Many fixes, pin properties, refactoring, device filters work with i/o pins. "PinnacleSys PCTV DVB-T tv tuner" initializes now. graphedt (DX SDK sample app) can now 'open' tv tuner and enumerate pins / communication / interface. DVBT Network Provider can now connect tv tuner. Fix crash calculating header location in KsGetDevice

BDA (Broadcast Driver Architecture) initialization fixes, debug traces added, work on filter functions, graphedit no longer shows dupe pins

PnP (Plug and Play) - fixes for Service names, fix device detection, store and update device resources properly, resource management

CmBatt - replace CmBatt driver (cmbatt.sys / compbatt.sys)

PCI - new PCI driver

PSDK - adding interfaces and definitions

NDK - add some IO functions

DirectX - device enumerator supports class enumerators

reactos-yarotows branch :
GDI: Use specific DC types instead of PVOID, let the drivers manage their resources
GDI: Continue fixing device context locking (mismatches and unneeded locking), respond to mode changes (system metrics), multi-threaded signaling

RosBE
Play a sound when build completes (in case you walk away or are paying attention to other things)

sir_richard passive-aggressive commit message counter: 4

edit: extend to 46658. I think I missed a pure "kernel" line somewhere...

[b4dc0d3r]

A different format this time. Bugs fixed (resolved since 1 April). Some other bugs were 'fixed' just by re-testing, I did not include those if they were very old because probably no one has been having the problem for a while.

3560 Control-Alt-Delete doesn't work if you kill all windows (handle keyboard when no windows are present)
5071 MSVC GDT macro/intrinsics (fixes crashes when compiling on MSVC)
2144 Task manager CPU graph assumes multiple-processor (or multicore), offers per-CPU graph option
4844 Cache username to reduce LSASS CPU usage, as in Windows
4067 'Format' command now checks if a drive exists before trying to format
4491 CMD.exe tab completion adds extra quotes
4235 Netmeter 1.1.3 crash
5263 Keyboard and mouse drivers loaded twice, visible in device manager
4661,4408 TaskMgr has had numerous problems in the process list - not showing some, not removing some, including system idle process. Appears fixed now. (still did not fix 1567 which includes 'access denied' message when killing the process)
4509 ECHO sends LF only, fixed to send windows style CRLF
4226 "Open With" dialog was not handling close/destroy messages properly, so only 'Cancel' would make it go away
1494 Log Off now asks for confirmation, including needed translation text

Still unfixed, but interesting development in FireFox 3 text rendering (bug 3481): FF3 uses low-level text output to be able to produce left-to-right output and other advanced fonts. This requires *Glyph* functions, not *Text* functions. A quirk: the developer of SmoothText also had problems with FireFox, like negative Y values and other oddities in calculating the width of text. There may be an off-by-29 issue in FireFox's width calculations (using the wrong character's width information), or it might be in ReactOS, or it might be the interaction of the two. Further investigation is under way. Interesting because some of these *Glyph* functions were fixed in order to solve Qt text layout problems, maybe a year ago.

KS: streaming statistics now available (tested on XP SP3)
ISAPNP driver removed because it just slows ROS down at this point, will be re-added if it becomes fixed/useful

arwinss :fix many winetest failures regarding 'last error' for user and threads, also leftover artifacts when moving/resizing app windows

[b4dc0d3r]

Some highlights:

46870 - some work on updating/saving themes, fix for bug 5016 "lost color scheme name in desk.cpl"

46867 - mjmartin continues USB work, mostly cleanup with some fixes and a few new functions.

Amine Khaldi continues to clean up headers on header-work branch

tkreuzer converts things from .def to .spec

46862 - mpiulachs releases C# version of rbuild (473 files!!!)

46861 fireball fixes race conditions by locking on timer routines on arwinss

Lots of stuff around -fno-leading-underscore switch, I'm hoping the next newsletter will explain a bit of this.

tkreuzer continues 64-bit work, on trunk and 64-bit branch.

46852 sir_richard is moving ABI around in the PnP manager, debug logs will suddenly have lots more information on where possible issues happen.

46851 fixes #4572, preventing empty message boxes on bad file type associations.

[jgwright]

This is very interesting, thanks.

[Aeneas]

Thank you from me, too. I will, however, generally keep out of this thread - this shall not become another usual discussion thread.

[unxed]

Just another "thank you" :)

[gabrielilardi]

You're doing a great job b4dc0d3r, people are getting what they were asking for, human readable information regarding reactos day-to-day work.
Keep it up!

[wildschwein]

I am so glad to here such a lot of good news !

So i am wondering if it perhaps at least slightly "smells" of ReactOS 0.3.12....
(The easter bunny didn't bring it , what a pity ! )

[Haos]

Perhaps this thread should be stickied?

[gabrielilardi]

Perhaps this thread should be stickied?

Nice idea, done.

[b4dc0d3r]

Catching up to 46903- with some hand-waving and vagaries since the last update

ARM headers get some attention

Bug 4383 - allow deleting multiple files from context menu (and a minor debug log fix for "SHFileOperation failed with [gibberish]" messages which now gives you an error number)
Bug 5300 - a contributed patch from Rafal Harabien: PING sends a packet, waits, sends, waits etc. so skip the last wait and it completes immediately. Eliminates the "You're done, stop doing stuff and let me type!" reaction from the user. Or was it just me who yells this?
Bug 4066 - a contributed patch from Rafal Harabien: Notepad's Load Settings function opens the registry, reads settings, loads the user's selected font, and closes the key. What's wrong with that? Well if the user doesn't have any settings saved (actually if the registry key does not open due to not existing or an error) no font gets loaded, so hoist font creation outside of the if() block and it executes regardless.
Bugs 5030,5116 - Desk.cpl - consistency and cleanup yields better results when selecting wallpapers (and highlighting the current setting)
Bug 4026 - restore code from r29143 (almost half the revisions ago!) where ShellExecuteA() executes the "open" action on everything instead of the default (it could be open, run, play, or whatever you have set as the default), and fix a hard-coded "Nix" string in the related error message. Why? "Control panels can be launched from run again". And they all rejoiced.
Some KS stuff I didn't look at this time around, it looked like prep work rather than anything users would notice, someone slap me with a trout if that's not true.
Bug 3573 - convert nulls to spaces when reading binary files, so now you can open files in notepad that you should be using a hex editor for. "But why would anyone do this?" you ask. Because sometimes you don't know if it's a binary file, and you'd still like to know what's in it.

Bug 4945 - r46875 - A previous commit r43772 removed the distinction between left and right ALT, SHIFT, and CONTROL keys, so now you get "A control key is down" instead of "The left control key is down." SetKeyState() was not informed of this at the time, so now it falls in line. I assume it's to allow programs who don't care which one is down to ask if one of the keys is pressed. GetKeyState() function allows you to ask "Is the left control down? What about the right control?" With virtual keys, you can just ask once for VK_SHIFT, instead of asking separately for (VK_LSHIFT or VK_RSHIFT). And same for CTRL and ALT.


Rbuild -
Fixes for recent gcc versions
Some fixes for RosBuild on MSVC

reactos-yarotows
After several tries, internal function GDIOBJ_LockMultipleObjs to prevent resource contention when locking multiple objects. Just like in SQL, lock things in the same order and you will never have a deadlock/race. Optimize by locking at most 3 objects at once, since 3 objects should be enough for anybody! Cleaning up BitBlt using this locking, and related work on NtGdiAlphaBlend and IntEngAlphaBlend

header-work
DDK/PSDK header fixes continuing

Edit: added information on 4945 for those unfamiliar with GetKeyState() API.

[livestrong2109]

Thanks for keeping everyone in the loop, its appreciated!
Go have fun with your girl
Enjoy Your Meal... Be sure to ask for dessert

[Angelus]

Thank you! I'm an avid reader of this thread. Keep on the good work!

[b4dc0d3r]

An interesting experiment. It takes a lot of thinking to understand the code changes, and I end up paraphrasing about half of the commit messsages because they are well written. I don't do any testing right now so I can't say this commit makes this type of error go away, only hint at what might be different.

I just went to type "CIA" into the address bar and realized it now feels like work. I'd much rather have a quick read of it during "brain rest" parts of the workday, instead of dedicating half an hour to bookmarking the current list and trying to communicate the significance of each change (to myself first, then to others).

I might stop completely, I might change style, I might turn into a purple butterfly. What I did accomplish is prove to myself how much effort it takes to complete a changelog, write a newsletter, or collate any third-party information into a semi-comprehensible jumble of letters. For now, you can read further posts by developing your extrasensory powers and reading my mind - it will save typing, and you can probably put it down on your personal development plan as training hours. If you succeed, please just focus on the ReactOS bits, the rest might not be suitable depending on your age and local laws and customs.

Experiment is still in progress, just developing more slowly. (Mods: if this is un-stickied at some point you won't hurt my feelings any so please do the needful)

[Haos]

I love reading CIA commit messages.