- Patch by Piotr Caban : Make it possible to activate a window with parent and no WS_CHILD flag in WS_NCLBUTTONDOWN function.
- Sync port from wine with modifications and addons.
svn path=/trunk/; revision=62610
Finally...
- Specify that we use MUTEXes instead of binary semaphores,
- Change two interface names in our code: ACPI_DEVICE_ID_LIST --> ACPI_PNP_DEVICE_ID_LIST and acpi_device_id --> acpi_pnp_device_id (only those names were changed in ACPICA).
- Implement AcpiOsPhysicalTableOverride which doesn't any table override (same behaviour as AcpiOsTableOverride).
- Latest ACPICA versions introduce the AcpiOsWaitEventsComplete API, which is used to wait for all asynchronous events to complete. But we do nothing at the moment ("inspired" from the behaviour from http://ftp.netbsd.org/pub/NetBSD/NetBSD-current/src/sys/dev/acpi/acpica/OsdSchedule.c , but I'm sure ReactOS ACPI specialists know better than me whether or not it is judicious to do so !!!)
- Extend AcpiOsRead/WriteMemory to be able to read/write 64-bit values, as required by the ACPICA implementation.
Fix the build. Now, only warnings from our code (not ACPICA) need to be fixed.
Real HW testers, please retest your real HW with this revision (or above) !!
CORE-8044 #comment Everything updated in revisions 62604 to 62608.
svn path=/trunk/; revision=62608
Addendum to revision 62606 : because of typedefs redefinitions (we, after, include acpi.h and co. together with DDK in other acpi bus drivers), we hackishly needed to disable them in actypes.h, and now we need to include ntddk *inside* acpi.h ...
Everything would be nice only if ACPICA would not have the "clever" (read: dumb) idea to define a ACPI_BIOS_ERROR macro, which is (you guessed it) already defined.... in the DDK via the standard bugcheck code macros.
So we introduce two hacks: the firs one right after the inclusion of ntddk.h *in* acpi.h to first undefine the macro (coming from bugcheck.h), *then*, we re-undefine it *after* the inclusion of acpi.h (in the precompiled header) so that drivers that need to use acpi (via the header) can after, include the ntddk.h ...
What a hack :DD
Part 2bis of our modifications to ACPICA code.
CORE-8044
svn path=/trunk/; revision=62607
Disable redundant typedefs that we already got from the inclusion of ntddk.h from elsewhere (and it was done like that in our old modifications of ACPICA).
Personally I don't like this way to go, but for the moment just apply something that (seemed to) worked before...
Part 2 of our modifications to ACPICA code.
CORE-8044
svn path=/trunk/; revision=62606
Do not assume that, when compiling from Windows, we cannot use GCC. In fact we can, and this is what we did before...
Move asm-specific code to where it should belong.
I've checked that: MSVC_asm_code(old_acpica_version) == MSVC_asm_code(new_acpica_version), so that I took its GCC equivalent from our old version of ACPICA and placed it there.
Part 1 of our modifications to ACPICA code.
CORE-8044
svn path=/trunk/; revision=62605
- Add an error message for Proprietary Software Execution Prevention
feature supported by recent CPUs.
- TODO: Implement this feature in the kernel.
svn path=/trunk/; revision=62598
ACHTUNG! ACHTUNG! Even if you can build ROS with MSVC 2008, you cannot boot it (see the below-mentioned JIRA report) ACHTUNG! ACHTUNG!
CORE-8023 #comment MSVC 2008 build resurrected in revision 62595.
svn path=/trunk/; revision=62595
For host-tools, including stdint.h when using MSVC is not reliable, so use a tried-&-tested solution (see mkshelllink & geninf tools): include stdint.h if we don't use MSVC, otherwise define just what's needed.
Another solution would be to create a include/host/stdint.h which does this same job, and in the host-tools, include this file.
CORE-8023
svn path=/trunk/; revision=62594
Move some defines where they belong (they are used in mkhive because they are employed in cmlib functions, and cmlib itself uses them too).
svn path=/trunk/; revision=62592
- Verify valid IRQL in KeDelayExecutionThread, just like in KeWait*
- Return from KeDelayExecutionThread after yielding execution
svn path=/trunk/; revision=62590
Reshuffle the test a bit so that it clearly fails on ReactOS:
- Gather creation timestamp right after the first creation and always compare to it. This is to workaround the move = copy + delete hack from ReactOS which is resetting the timestamps (and thus making the test pass ;-)).
- Wait more than 20ms (1s). If Microsoft FastFAT has a creation resolution of 10ms, we don't. So it would hide the thing as well.
With such changes, the test still works on Windows, whereas it fails on ReactOS
CORE-7272
svn path=/trunk/; revision=62589
Extend the maximum password length to 127 characters for unattended setup too.
CORE-7762 #comment Fixed it! Thanks a lot!
svn path=/trunk/; revision=62580
Add a usermode test application for tunnel cache.
It is based on MS article at: http://support.microsoft.com/?kbid=172190
Tested successfully on Windows 7 & NTFS. Feel free to test on w2k3 NTFS or FAT/FAT32.
svn path=/trunk/; revision=62579
Add a note for translators to keep the administrator account name string resources in syssetup.dll and samsrv.dll synchonized.
svn path=/trunk/; revision=62578
- Store the administrator account name as the default user name.
- Extend the maximum password length to 127 characters.
CORE-7762 #resolve
svn path=/trunk/; revision=62577
- Mark Str* functions as private in comctl32, they should be imported from shlwapi instead
- Remove the associated hack in rapps
CORE-7786 #resolve
svn path=/trunk/; revision=62570
- Remove unnecessary use of allow_warnings
- Do not hide warnings that indicate actual problems in setuapi (non-Wine test, should move to apitests) and winmm (Wine bugs, fixed upstream)
svn path=/trunk/; revision=62569
- Remove unnecessary allow_warnings throughout the code base
- Make -Wnarrowing an error unless excepted specifically (only done in explorer)
svn path=/trunk/; revision=62566
try_return() == try_return(S) with nothing in S . The code sometimes use it.
Shut up MSVC warning C4003: not enough actual parameters for macro 'try_return'.
svn path=/trunk/; revision=62562
- Partly update/sync Windows-MSVC.cmake with RosBE 2.1.1 : few flags "fixes". Amine, can you please check whether there are other things to sync too (maybe those WinCE things, even if it's unuseful for ROS itself, and maybe some other vs. MSVC_C_ARCHITECTURE_ID _MSVC_C_ARCHITECTURE_FAMILY flags, and so on... and with syncs of other .cmake files??)
- Remove unneeded manually setting of stack size, it is done automatically by the compiler.
- Add a "/MANIFEST:NO" switch to the linker so that (unuseful) manifest files are not automatically created and added to the host-tools, especially, because that causes some concurrent accesses failures on MSVC 2005/2008...
CORE-8023
svn path=/trunk/; revision=62560
- GCC: Cosmetic changes only.
- MSVC:
* Cosmetic changes (the same as in GCC).
* Only VS 10+ resource compiler supports /nologo , therefore define a rc_nologo_flag that sets /nologo for VS 10+ rc.exe, and nothing for VS < 10.
CORE-6997 #comment Only VS10+ rc.exe supports /nologo, modify the fix in revision 62559.
CORE-8023
svn path=/trunk/; revision=62559
Set PROMPT to $P$G$I instead of $I$P$G, so that the dumb Win cmd.exe can parse "completely" the prompt string (and succeeds in displaying the $p$g part) before failing on the $i part...
See rev.59837 and
CORE-7634 #comment A better fix was committed in revision 62557.
svn path=/trunk/; revision=62557
