With this attempt Thomas proved that it is possible to use patchbot for both trunk and rostests patching. Ask him how he did it:>
On Wednesday, October 05, 2011 9:05 AM, ReactOS.Bugzilla(a)reactos.org wrote:
> --- Comment #1 from ThFabba <thfabba(a)gmx.de> 2011-10-05 09:05:44 CET ---
> Created an attachment (id=6782)
> --> (http://www.reactos.org/bugzilla/attachment.cgi?id=6782)
> patch 2
>
> Maybe this works :p
>
--
With best regards
Caemyr
Hi,
I've been looking at VirtualBox 4.1.x guest additions, which don't install
in ROS [1] and came up with a patch [2] that seems to fix them.
While patchbot says this is okay [3] (and the last hunk even fixes a few
setupapi tests), I am very unsure about the SP_COPY_NOOVERWRITE part.
The correct solution might instead be to handle SetupCopyOEMInf returning
ERROR_FILE_EXISTS or I might be "fixing" something in the completely
wrong place.
It would be great if someone with some insight into setupapi
(Eric? Hervé? ;]) could review this and give some pointers.
Thanks!
-Thomas
[1] http://www.reactos.org/bugzilla/show_bug.cgi?id=6522
[2] http://www.reactos.org/bugzilla/attachment.cgi?id=6753
[3] http://www.reactos.org/testman/compare.php?ids=8272,8294,8303,8305
Patch explanation:
- The first hunk isn't really required, but I guess it shouldn't break
anything, and seems to find what it's looking for this way.
- The second hunk is in the code path used by VBox guest additions setup.
It is apparently called with an already-existing inf file, but I have
no idea why. :|
- The third hunk is a trivial bug that our CreateService just didn't
complain about before r53872/r53886
Though the August minutes were already posted on the forum, they are being
resent here for people who may have missed them.
August Summary
A proposal was made to do winesyncs before a release to force the project to
do a better job keeping things in sync and perhaps do a better job making
syncs feasible/less time consuming. Rejected by plurality of votes.
Questions about how/whether to continue the driver signing program was
raised. An advisory vote was held and the majority wished to continue it in
one form or another. Feedback based on the advisory vote was incorporated
into revised guidelines posted on the Driver Signing wiki page. New projects
are now required to find a sponsor ReactOS "core developer" willing to audit
their code if they wish to get their driver signed.
Decision to integrate arwinss into trunk under a separate directory
structure and make it a compile-time option. Passed by majority vote.
Decision to switch main Linux build server to use cmake. Passed by majority
vote.
September Summary
Agenda
1. CMake Adoption Status
2. Release Preparation, Determination of Release Date
3. Webportal Upgrade Status
4. ReactOS Target Version
Report
1. The build environments for both Windows and Unix have gone through
several release candidates and are effectively finalized. For the Windows
BE an installer release client using MSI instead of NSIS also needs to be
prepared. Minor problems still seem to exist on Mac OS X when using a
combination of GCC and LLVM, but these were deemed non-blockers. Building
ReactOS using CMake has been reported to take more time than using rbuild,
though developers are still working to verify and determine the cause.
Current theory is related to changed debug format and more thorough
dependency checking by CMake. At present the plan is to put the build
environment release candidates on the Linux build servers and determine if
there are any problems. The Windows build servers have already been running
the new build environments for testing purposes.
2. Due to the significant number of regressions and overall bugginess of
ReactOS' current state, release scheduling has been deferred until the next
meeting. A list of regressions and bugs has already been compiled and the
team intends to work its way through the reported issues in the meantime.
At the next meeting, the OS' state will be re-assessed to see whether a
release is viable or not.
3. Most of the modules needed to bridge Drupal with things like Bugzilla
have already been completed. To speed the transition, a decision was made
to retain the current website design. Once the transition is complete, the
web team can go back to updating the site's theme. Desired content changes
will also not block transition to the new backend. No timetable was given
for when the transition would happen.
4. For a while the project adopted the goal of targeting the Windows 2003
kernel and the latest version of Windows for user mode components. Several
developers however raised objections to this and presented examples where
this methodology does not work. A specific example involved symbolic link
support, which required Microsoft rearchitecture not only NTFS but also the
kernel's security systems. Its addition in ReactOS, assuming it is done
correctly, would require changes to the kernel to adopt features from the
Vista kernel, thus violating the kernel target. With these considerations,
the project has decided to set both the user and kernel targets to Windows
2003. Dependency on Wine for many user mode libraries actually complicates
this, as Wine has been adding Vista and later APIs without proper
segregation and the developers will need to figure out how to deal with this
in the long run.
Am 02.10.2011 22:59, schrieb akhaldi(a)svn.reactos.org:
> Author: akhaldi
> Date: Sun Oct 2 20:59:15 2011
> New Revision: 53941
>
> URL: http://svn.reactos.org/svn/reactos?rev=53941&view=rev
> Log:
> [RBUILD]
> * Plug a leak.
Bugfix of the month! ;-)
Hi,
at present I work for mid-sized company for ticketing, parking and mobile
solutions in Germany (and Great Britain). I personally work in a project for
railway services (using gsm-r network). To understand the tools given below I
want to figure out our target and environment in short. We develop for
embedded Linux systems with ARM-CPU and Qt-Framework. Therefore C++ was and
is used as programming language.
Apart from he usual tools like
- make-alike build tools (written in perl - a project specific one)
- regular make files for some special targets
- SVN for source code management
we use some interesting tools to keep the source readable and to preserve a
minimum of source quality automatically (bear in mind we use C++ instead of
C, for our purpose we need to alter some tools for our needs):
- Hudson for automatic build and checking the source (see below)
- cpplint.py (by Google) for some checking the source
- cppcheck for another check of the source
- doxygen for checking the missing documentation (in that special case check
only by tricky configuration instead of costly generating all documentation
files)
- format the source code (after agreement of all devs) in a uniform style with
astyle before commit (for ReactOS better as commit hook)
- finally we gather all check results in one table and mark the results for
hudson as stable (if no errors), failed (in case of any errors), unstable (in
case of some low-prio warnings)
Maybe we can think about some of them, because even if some problems of
development remains, it might give us some edge and may also attract some
devs. I don't make any suggestions, but I want to tell you how others work.
Matthias
PS: We are the first project in the company going that way and other project
groups would like to adapt the way, because they see the benefit for
their/other/future projects.
--
Matthias Kupfer phone +49 371 236 46 52
Wilhelm-Firl-Straße 21 mobile +49 160 859 43 54
09122 Chemnitz, Germany