- Uniform function declarations in .c and .h with fixes and SAL2 improvements.
- Use LDR_[UN]LOCK_LOADER_LOCK_FLAG_XXX correctly.
- Fix incorrect UsedSize calculation in LdrQueryProcessModuleInformationEx (caught by ThFabba).
- Whatever the callback request stop the enumeration or not in LdrEnumerateLoadedModules, the following operations should be the same.
- Fix 2 incorrect DPRINT1 that printed incorrect parameter.
- Return error if RtlAllocateHeap failed in LdrpGetProcedureAddress, and add comments about NT6.2 new changes.
Don't display applications that don't exist. In Windows XP, Exe files that no longer exist are hidden but their registry entry is not deleted.
JIRA issue: CORE-19117
Display the class icons for the installed network components.
Also fixes a bug where the selection change whacked the checkbox of the first item.
Use 0x7fffffff to append the item at the end of the list, as optimization.
- Handling ADN_DEL_IF_EMPTY is critical to avoid data loss.
- SetFileAttributesW failing is not fatal, the delete might still succeed.
NOTE: Not in Wine yet.
- Fix TCHAR/WCHAR mis-usage,
- Fix as a result, a buffer overflow (GlobalAlloc takes the size in
bytes, but a number of characters was passed to it instead).
- Remove usage of unsafe string function. Now the item text is directly
retrieved within the allocated buffer.
- Adds support for registry controlled menu separators and the documented
values to turn off verbs.
- Adds support for CMF_OPTIMIZEFORINVOKE, CMF_NODEFAULT, CMF_DONOTPICKDEFAULT,
CMF_EXPLORE and CMF_DISABLEDVERBS.
Bugs fixed:
- A verb with "Extended" set in the registry could cause the menu to invoke
the incorrect command! This happened because skipping InsertMenuItemW
caused InvokeCommand to use the wrong index with m_StaticEntries.
- Uses IS_INTRESOURCE instead of HIWORD to check if something is a string
(only matters on 64-bit).
- TryToBrowse leaking a PIDL when calling ILCombine.
Notes:
- This PR introduces the RosGetProcessEffectiveVersion() helper function
discussed in chat.
- Relaxed FAILED_UNEXPECTEDLY to FAILED in two places because IContextMenu
cannot assume that it has a site that leads to IShellBrowser.
This adds support for more of the FOLDERSETTINGS FOLDERFLAGS.
- Turned on LVS_EX_DOUBLEBUFFER even though there is no setting for it.
XP has it on for folders but not the desktop and this matches that.
Vista might turn it on for its shadow labels on the desktop, further
investigation required.
- Fixes a bug where LVS_EX_SNAPTOGRID was incorrectly assigned where
the WS_EX flags go.
- LVS_AUTOARRANGE seems to be incorrectly forced on but for compatibility
I decided to leave it but added a comment. The existing code already
has the check for FWF_AUTOARRANGE.
This finishes the 640x480-ification within [NETCFGX]. All dlgs addressed now.
- Allows the final 3 property sheets IDD_TCPIP_ADVIP_DLG, IDD_TCPIP_ADVDNS_DLG and IDD_TCPIP_ADVOPT_DLG to be displayed fine in 640x480 screen resolution by limiting their dlg unit height to 220
- tweaks a bit of the controls vertical positions to look better, e.g. LTEXT "interface metric:" was moved one pix upwards relative to IDC_METRIC to make that look more vertically centered
- IDC_AUTOMETRIC checkbox was moved slightly upwards in relation to the groupbox it was embedded into, which does it make look slightly better
- width of IDC_SUFFIX was enlarged for all languages to the value that was historically chosen for uk-UA.rc. (Only ru-RU.rc continues to do that differently, as it uses even longer text and therefore has to shift 2 controls towards the left to make that fit into the dlg. That part was there also before this PR)
Main goal was making the dlg display well in 640x480 resolution.
Since the dialogs in that propsheet had very different sizes between the languages
that was a lot of work. Now all of the propsheet-dlgs are harmonized in size to 252x220.
Also I found a way to align some controls in a way that they are more immune
to differing languages lengths for specific strings, so we can currently
to 100% use the same controls dimensions and positions also,
which is a very nice trait. I double-checked all dialogs visually in ResHacker
for all languages (except no-NO.rc for technical reasons), and they do look all
very well like that. The slider for the alarms dlg I enlarged in x,
it has 100 distinct positions, so a larger slider helps the user
not having to position the mouse extremely accurate.
I also fixed various typos, and several places where text cutoff occurred for many languages,
too many to name them all.
Some dummy-strings "unknown" I changed to "?" instead.
That still allows to see its position in a previewer,
but does not lure all the translators into translating that nonsense,
and therefore keeps the rcs better in sync. It also shrinks binary size.
Both dialogs can easily fit 640x480, like their twins from XPSP3 and 2k3sp2
For IDD_TCPIP_BASIC_DLG we can achieve that by just shrinking the upper groupbox in height by 10 units, and then move all controls below that by 10 units upwards. That doesn't require much brain, only diligence.
For IDD_TCPIP_ALTCF_DLG it is even easier, because the dlgs contents are small enough yet to still fit 640x480, so only the dlg itself must be shrinked for now. Later if more stuff will be added, we can perform pretty easy still by just moving everything a bit upwards. But we can leave that for later.
Fix a common typo in a 3rd dlg regarding a missing space before "(in order)". I decided to put that into this PR, because I didn't touch that dlgs dimensions yet. So the change is not mixed with position changes and therefore the review remains simple.
For ja-JP.rc fix a
-FONT 8, "MS Shell Dlg"
to
+FONT 9, "MS UI Gothic"
- adds 3 missing accelerators for zh-CN.rc, zh-HK.rc, zh-TW.rc for IDC_SHOWTASKBAR
- shrinks the dialog in y to fit on a 640x480 screen, like the XPSP3 and 2k3sp2 dlgs do
- for that some controls positions have to be slightly moved up (which cannot lead to text truncation here, guaranteed)
- only 3 controls change in size:
- IDC_COMPONENTSLIST gets 3 units smaller in height, but is still larger than in XPSP3/2k3sp2. So it is still large enough
- IDC_NETCARDNAME shrinks a bit in width, but is still larger than in XPSP3/2k3sp2, so what fits there will also fit for us
- pushbutton IDC_CONFIGURE is slightly increased in width for all languages because that looks better
except ru-RU.rc (which had the largest width) it is slightly smaller now, but still large enough.
We can use the same width here now for all rcs!
- strip the lorem ipsum text from IDC_DESCRIPTION, which is never supposed to be displayed. That syncs all rcs in that line and slightly shrinks the binary size and fixes CORE-19284.
- add a FIXME comment for a missing accelerator in sv-SE.rc
*as a consequence all rcs do now use the exact same positions and sizes for all controls again, YIPPIE!*
The only exception that the dlg had before from that rule, was the ru-RU.rc IDC_CONFIGURE was larger in x.
But now that is not needed anymore.
George explained that this is supposed to fail like that,
aka "works as designed". So let's at least mute this now,
as it is spammed with every taskmgr refresh cycle
and allows me to reduce logfile-sizes from ~50MB back to ~2MB
when building ros on ros.
err:(dll/win32/advapi32/wine/security.c:309) NtOpenProcessToken failed! Status c0000022.
Fixes CORE-18521
Also strip some EOL-whitespace in the file, and slightly improve the header.
Based on KRosUser's aclui.patch. The list view will automatically delete the image list.
Don't delete the image list in DestroySecurityPage function.
CORE-19187
On Windows the Network connections folder uses a binary value
and someone forgot about Little Endian when converting it to a DWORD.
Addendum to 7eb3fcf1d (r66485). CORE-9276
This issue has not been very visible because of bugs in CRegFolderEnum.
and move debug.h after all includes. Addendum to 60b0afc3a (PR #5818)
dsound_new: Addendum to 5974fe1 (r45584).
hdaudbus: Addendum to cf7fc81 (r68311).
Based on KRosUser's patches.
- Add missing version strings "Comments" and
"LegalTrademarks".
- Add version strings "PrivateBuild" and
"SpecialBuild" if necessary.
- Trim the version string for pretty display.
CORE-19270, CORE-19114, CORE-19275
This fixes icons that should be drawn transparently not showing as such until window gets repainted.
For example, hidden shell folders that are not drawn transparently until the filebrowser window is resized...
CORE-16722
- Introduce AddScreenSaver[Item] helpers for adding screensavers in
the list (either given by file path or by item structure).
- Rename "AddScreenSavers" to "EnumScreenSavers", and
"ScreenSaverItem" type to "SCREEN_SAVER_ITEM".
- Ensure the stored "SCRNSAVE.EXE" value is NULL-terminated.
- Add the current screensaver, specified by the "SCRNSAVE.EXE" registry
value, in the list if it isn't already present; otherwise select it
in the list.
- Make the filename comparison case-insensitive.
CORE-6812 ; see also commit ff0951356 (r66688)
This function "just" changes the per-user SCRNSAVE.EXE registry value
to point to the new specified file, changes the SPI values and opens
the desk.cpl "Screensaver" property tab for letting the user change
the new selected screensaver properties.
In particular, it does *NOT* copy the specified file to System32 or
anything else (and doesn't verify that it is a valid PE executable).
Our previous implementation did none of that, and was also relying on
some private setupapi functions.
We now behave closer to Windows' desk.cpl.
Additionally:
- ReactOS-specific feature (compile-time define) disabled by default:
Verify that the specified file actually exists, before changing
the screensaver.
- Use NDEBUG, disabling DPRINT by default;
- Improve InstallScreenSaver[A,W] spec entries;
- Remove NUM_APPLETS define and use _countof() instead.
- Only export stdcall annotated functions on x86
- Turn annotated stubs / forwarders into normal exports
- Note: MSVC removes decorated exports / forwarders, as soon as the undecorated version is exported, even when _declspec(dllexport) is used, no idea how to fix that.
This avoids having desk.cpl depend on a private function that may change
or disappear, and increase the probability of being able to use that CPL
across different Windows versions.
(Note: this pSetupMultiByteToUnicode was one of those whose name changed
between Windows 2000 and XP+)
Addendum to commit 8c4b0c914.
Base ourselves on pSetupIsUserAdmin() and other similar functions in our
codebase. Note that what we are actually interested here, is whether the
current thread runs with Administrator privileges.
(As noticed by contributor 'whindsaks', "Not only is this code simpler,
it now will correctly handle deny-only SIDs in the token!")
This mostly reverts commit 221f21c62 (r29328), that was a hack for
building with MSVC. We can nowadays correctly link with the shell32
SHCreatePropSheetExtArrayEx() export.
- Convert some notations to SAL2
- "Name" parameter in LdrGetProcedureAddress/LdrpGetProcedureAddress should be optional
- "DllCharacteristics" parameter in LdrGetDllHandle should be optional
- [EXPLORER] Set maximum balloon width for notification area
- [COMCTL32] Fix balloon tooltip shape
- Limit balloon tooltips from extending past the edges of the monitor, not the edges of the work area.
- Instead of simply repositioning the main rectangle when the balloon is too far towards one edge of the screen,
try flipping the balloon the other way. This is the same behavior as Windows Server 2003.
- Tweak some values used to draw the balloon tooltips to more closely follow the Windows balloon tooltip style.
- Removing trailing whitespace.
- While the new changes are guarded, consider cancelling our Wine sync for common controls.
Our common controls are responsible for many graphical issues and lack of features throughout ReactOS.
CORE-19109
Adds the option to set the address edit box to use the display name or the full path. Also refreshes the window title and edit box in all open explorer windows when changing these settings using the folder options dialog.
## PROPOSED CHANGES ##
- Create a new CabinetStateSettings type that inherits from the CABINETSTATE type. This allows us to add additional cabinet state settings not exposed in the CABINETSTATE type as well as adding a Load() method to easily populate the cabinet state settings.
- Add a global cabinet state settings object. While most settings in browseui are stored independently in each shellbrowser window, cabinet state settings are global and apply to every shellbrowser window. This can be confirmed on Windows Server 2003 and Windows 7.
- When receiving the WM_SETTINGCHANGE window message from the folder options dialog, refresh the title of the window and the text in the address edit box. This is the same behavior as Windows Server 2003 and Windows 7.
Add a DWORD registry value to HKCU\...\Explorer\CabinetState\FullPathAddress to allow users to toggle this setting on or off in our folder options.
CORE-9277
We do use the birth year 1996 instead of 1998
within our msgina pictures already,
and also within the [SDK] for autogenerated file-info.
Since SHELL32 IDD_ABOUT is used by many applications it makes sense to harmonize with that.
Was done already during 0.4.15-dev for SHELL32, but we have to do it again,
since some recent translation PRs to match the en-US.rc again.
fixes CORE-18967 'Taskbar toolbar - Right click > Open folder fails'
by restoring the code-path we removed with the guilty
0.4.15-dev-3847-g /9b716539182864bc077e362febcda74da50287ed
and add test_sei_lpIDList to shell32:ShellExecuteEx apitest test to protect
that functionality on future attempts to simplify.
movaps is a 128 bit aligned move, we need a 64 bit unaligned move, so use movsd.
Fixes a crash in oleaut32_winetest tmarshal.
Also add a comment about the use of movd instead of movq.
- CDM_SETDEFEXT message handling was a
missing feature of COMDLG32.
- Make info->defext dynamic allocation.
- On CDM_SETDEFEXT handling, replace
info->defext data.
CORE-15020
- comdlg32:filedlg could encounter an assertion error in CWindow::SendMessage.
- Assert the existence of the listview control m_ListView in the CDefView::LV_... function.
- Actually check the existence of m_ListView outside the CDefView::LV_... function.
ROSTESTS-388
Follow-up to #5634.
- Implement CMruBase::_UseEmptySlot.
- Implement CMruLongList and CMruShortList.
- Add CMruClassFactory class and modify
DllGetClassObject function by using it.
- Add shdocvw_apitest.exe.
CORE-9283
Fixes failing opengl32:sw_extensions test.
The test expects version `1.1.0` which is what Windows reports,
and ReactOS Mesa3D was reporting just `1.1`.
- Implement CShellBrowser::GetPropertyBag
a little by using SHGetViewStatePropertyBag.
- Add SHGetPathFromIDListWrapW prototype to
<shlwapi_undoc.h>.
CORE-9283
This is prerequisite to supporting Vista+ iteration of the msstyles format,
and also for implementing GetThemeStream() later.
Visual styles compiled for Vista+ can now use PNG images in the IMAGE group.
This does not affect the ability to use BMP images, they can be mixed now.
CORE-18958
Introduce an emergency log-off and restart feature, available when holding
down the control key and selecting "Log Off" or "Shut down" in the MSGINA
Ctrl-Alt-Del security dialog, similar to Windows Server 2003 and newer
Windows versions.
Follow-up to #5610.
- Add _EnsureWriteBag, _PruneMRUTree, _GetMRUSize and
_GetMRUSlots helper functions.
- Add code into _GetMRUSlot, _FindNearestInheritBag, and Write.
- Add CLSID_MruPidlList and IID_IMruPidlList definitions into
<shlguid_undoc.h>.
- Add IMruPidlList interface into <shlobj_undoc.h>.
CORE-9283
- Don't forget to release RTL_RELATIVE_NAME_U before return
- Ignore lpSearchFilter correctly as on Windows
Fixes 6 failures in kernel32:FindFiles test in LiveCD.
Previously it was in cpprt, which is a support library for C++, containing functions that are not exported by msvcrt. But since Vista __CxxFrameHandler3 is exported by msvcrt. Therefore move it to crt, and to satisfy pre-Vista configurations, also add it to msvcrtex.
Follow-up to #5590.
- Add CViewStatePropertyBag class.
- Add SHGetViewStatePropertyBag definition.
- Add FreeViewStatePropertyBagCache function and
use it in DllMain to free the cache.
CORE-9283
Currently ReactOS' winsock2 implementation lacks of non-blocking sockets support for recv() apicall, this causes that applications that make use of this feature can lead to unexpected behaviors, one of them is Nginx web server, which uses non-blocking sockets when serving pages through Https protocol.
CORE-14486
It also brings us significantly closer in master head to running Firefox 52, Mypal 29.3.0 and New Moon 28 browser, where the latter allows to connect to mattermost from within ReactOS.
In master head an additional reg file is needed to stop us from exporting specific NT6+ APIs, but in older releases all that should work out of the box with this brilliant patch.
Co-authored-by: Julio Carchi Ruiz <julcar@informaticos.com>
Co-authored-by: Stanislav Motylkov <x86corez@gmail.com>
This works similarly to how our shell stores its settings from a software design standpoint.
- Add settings.cpp file and ShellSettings structure to load and save settings.
- Add a registry value to hivedef.inf for the locked toolbar state.
This prevents a bug where the associated registry key cannot be opened or saved to.
- Add new BWM_SETTINGCHANGE window message to refresh the UI on setting changes
and send it to every open window when saving settings to the registry.
- Add new BWM_GETSETTINGSPTR window message to share the shellbrowser settings
structure pointer with child windows and toolbars.
When creating a new process the monster function CreateProcessInternalW calls BaseInitializeContext to set up the context for the initial thread. One of the parameters is the PEB pointer. Instead of passing the pointer to the newly created process' PEB, the function was passing it's own PEB address, leading to test failures. How this is not an issue otherwise is a mystery to be resolved by someone else.
Also remove ros_skip_flaky marks in kernel32_winetest:process (yes, flaky, because sometimes the PEB addresses actually match)
Create kernel32_vista_static library and link both kernel32_vista and kernel32 to it.
Export some vista functions from kernel32.
Export BaseProcessInitPostImport for Vista+, too, because ntdll needs to dynamically link to it.
Make sure RtlRemovePrivileges gets compiled, when DLL_EXPORT_VERSION is 0x600+
Enable all existing functions from rtl_vista.
Hack CsrNewThread export to exist on later versions, too
The infoPtr->marqueeRect structure is not currently initialized before the
marquee highlight sequence starts, resulting in the RECT having initial
coordinates of (0,0)-(0,0). These coordinates cause the first item in
the listview control to be identified as being within the range of the
marqueeRect's coordinates.
That item is then set to LVIS_SELECTED even though it is not part of the
marquee selection.
Wine commit: daf95aaadf3a59f0ccc129a853327417b5e4f07c
author: Hugh McMaster <hugh.mcmaster@outlook.com>
Mon, 26 Apr 2021 10:59:51 +0000 (20:59 +1000)
Follow-up to #5531. These two functions are necessary for INI file property bag support.
- Implement SHGetIniStringW and SHSetIniStringW.
- Strengthen SHPropertyBag testcase of shlwapi_apitest.
CORE-9283
- If the source and destination of copying file are on the same directory,
then enable FOF_RENAMEONCOLLISION flag. End if.
- Don't show error message IDS_COPYERRORSUBFOLDER or
IDS_MOVEERRORSUBFOLDER if the source and the destination are the same.
CORE-18599
Follow-up to #5511.
- Add CRegPropertyBag class.
- Implement SHCreatePropertyBagOnRegKey function by using CRegPropertyBag.
- Strengthen SHPropertyBag testcase in shlwapi_apitest.
CORE-9283
- Implement SheRemoveQuotesA and SheRemoveQuotesW functions.
- Add She testcase into shell32_apitest.
- Add SheRemoveQuotesA/W into <undocshell.h>.
CORE-9277
The previous version resolved the path of the parent then did logic
assuming simple pidls. This now resolves each source directly.
This change addresses a longstanding TODO in copying and moving to shell folders.
It's a simple change, but it removes a bit of code and makes things simpler.
Corresponding Wine PR: https://gitlab.winehq.org/wine/wine/-/merge_requests/3360
SHLimitInputCombo function sets limits on valid characters for a combobox control.
This function works like SHLimitInputEdit, but the target is a combobox instead of a textbox.
CORE-9277
- _WIN32_WINNT had wrong value to do workaround on Vista+ RegGetValueW.
- Do #undef and #define _WIN32_WINNT _WIN32_WINNT_VISTA in assoc.c.
- Delete _WIN32_WINNT hack in CMakeLists.txt.
CORE-9283
- Implement SHPropertyBag_ReadBSTR, SHPropertyBag_ReadStr, SHPropertyBag_ReadPOINTL,
SHPropertyBag_ReadPOINTS, and SHPropertyBag_ReadRECTL functions.
- Add link to oleaut32 in shlwapi_apitest.
CORE-9283
Add a string pointer local variable, assign it and operate on it.
This should eliminate an exception and make LsapIsLocalComputer()
function operate correctly.
CORE-18996
- Add VariantChangeTypeForRead helper function.
- Implement SHPropertyBag_ReadType, SHPropertyBag_ReadBOOL, SHPropertyBag_ReadBOOLOld, SHPropertyBag_ReadSHORT,
SHPropertyBag_ReadLONG, and SHPropertyBag_ReadDWORD functions.
- SHPropertyBag_ReadInt is an alias to SHPropertyBag_ReadLONG.
CORE-9283
- Implement SHPropertyBag_Delete, SHPropertyBag_WritePOINTL, SHPropertyBag_WritePOINTS, and SHPropertyBag_WriteRECTL functions.
- SHPropertyBag_WriteInt is an alias to SHPropertyBag_WriteLONG.
- Modify shlwapi.spec.
- Modify shlwapi_undoc.h.
- Add SHPropertyBag testcase to shlwapi_apitest.exe.
CORE-9283
- Set `HKCU\SOFTWARE\Microsoft\Windows\CurrentVersion\Explorer\CabinetState\FullPath` to 0 by default.
Adding this registry value also enables the "Display the full path in the title bar" option in our folder options window.
- Query the same registry key when the file explorer is started to determine whether to use the full path or the display name in the window title.
CORE-19046 CORE-18904
- Add Internet icon on Desktop. You can hide/show
the icon from Desktop's customization.
- Modify "HKCR\CLSID\%CLSID_Internet%" registry key.
- Add IDS_INTERNET and IDS_INTERNET_DESCRITION into ieframe.dll.
- Modify folders/CDesktopFolder.cpp and folders/CRegFolder.cpp to add the icon.
CORE-18625
Co-authored-by: Stanislav Motylkov <x86corez@gmail.com>
Co-authored-by: Joachim Henze <joachim.henze@reactos.org>
This fixes 'Attempt to free an invalid HEAP address' error. CORE-18841
Confirmed by running new tests written by Mark Jansen, showing that
previous behavior was wrong. While here, also fix a typographical error.
New ReactOS can zip/unzip files and folders in UTF-8 filenames.
You can also choose the codepage of filenames to zip/unzip via
the ZipCodePage / UnZipCodePage registry values on the registry
key HKEY_CURRENT_USER\Software\ReactOS.
Windows 8 or later also support UTF-8 zipped folders.
You can also use 3rd party software to zip/unzip in older Windows.
- Use <atlconv.h> for string conversion.
- Use zipOpenNewFileInZip4_64 instead of zipOpenNewFileInZip3_64,
and then add MINIZIP_UTF8_FLAG flag.
- Set the filenames in UTF-8 by using CP_UTF8 codepage.
- Codepage is user selectable via registry settings.
CORE-16668
- Add _ILGetDriveType helper function.
- Use _ILGetDrive and GetDriveType functions to determine the drive type.
- If it was a CD-ROM drive, then remove SFGAO_CANRENAME flag.
CORE-18272
- Extend some Ansi strings to Wide strings.
- Check the UTF-8 flag (1 << 11). If UTF-8, then use CP_UTF8.
- s/LPCWSTR/PCWSTR/.
- s/LPWSTR/PWSTR/.
CORE-16668
- Use proper window styles when creating status bar window.
- Specify WS_VISIBLE style only when status bar display is enabled.
- Use ::CreateWindowExW() with a global namespace scope operator.
- Fix some magic constants.
CORE-19027
All ordinals > 128 regressed by
0.4.15-dev-305-g dd7fb63cd1
and were off-by-one compared to 2k3sp2 and the ros-before-state.
Now all 684 ordinals seem to be fine again.
Adding the stub might cost a few bytes unfortunately.
We do use the birth year 1996 instead of 1998
within our msgina pictures already,
and also within the [SDK] for autogenerated file-info.
Since SHELL32 IDD_ABOUT is used by many applications it makes sense to harmonize with that.
CORE-14015, CORE-6942
This makes Windows 2003 CMD.EXE to start batch files again.
Addendum to commit 17d42ae2a (r58868).
Replaces PR #3084 by freely importing and adapting Wine patch
17110a0a89
```
advapi32: Improve the SaferComputeTokenFromLevel stub.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=47274
Signed-off-by: Hans Leidekker <hans@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
```
Only done in our own code, not in 3rd-party or in PSDK headers
(official ones also have such typos in defines...)
(NOTE: apparent "Uknown" typo in dll/cpl/sysdm/smbios.c is on-purpose:
typo by the manufacturer.)
- Convert all modeless dialog boxes back to modal dialog boxes.
- Use WM_TIMER to detect SHIFT key press instead, as you can't
directly intercept messages on modal dialog boxes.
- Remove some unused functions.
Fixes explorer crash on shutdown cancel. CORE-17749
Actually I was working on a much bigger commit in the same context for releases/0.4.7
when I realized that related glitches do also affect master head
and also other release-branches in some areas.
- Selecting all in system32 was too slow.
- Make CDefView::UpdateStatusbar function asynchronized by using a thread.
- It makes selecting all very quick.
- Add m_hUpdateStatusbarThread to asynchronize.
CORE-18663
- FCIDM_SHVIEW_NEW resource string had some problems in localization and keyboard usability.
- Add IDS_NEWITEMFORMAT resource string ("New %s") to format the new file item string.
- Modify CNewMenu::NewItemByNonCommand for IDS_NEWITEMFORMAT.
- FCIDM_SHVIEW_NEW is provided for [New] menu item, so it should have an ampersand.
Add a comment for it and then retry the translation.
CORE-18706
In C, an inequality comparison between a signed and an unsigned integer
ends up with an unsigned comparison.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=52628
Wine commit: 0de8d01b09b1cc7ca34f7ae3890b4a416ff801fe
author: Eric Pouech <eric.pouech@gmail.com>
Wed, 30 Mar 2022 12:57:44 +0000 (14:57 +0200)
committer: Alexandre Julliard <julliard@winehq.org>
Fri, 1 Apr 2022 20:59:40 +0000 (22:59 +0200)
Note that since str is a buffer it cannot be NULL.
Based on Wine commit: 6a4347be371bea544e0d4a81f4cfacd29af18d6e
author: Francois Gouget <fgouget@free.fr>
Mon, 16 Sep 2019 14:01:04 +0000 (16:01 +0200)
committer: Alexandre Julliard <julliard@winehq.org>
Mon, 16 Sep 2019 20:58:42 +0000 (22:58 +0200)
- Strip inconsistent spaces from the spec file.
- iphlpapi_main: Harmonize inconsistent and line-wasting function-comments and fix mixed space-tab-indentation.
- ipstats_reactos.c: Compactify horizontally accordingly to the coding guidelines
I am interested in that code for a backport, so I want it to be well-formatted.
Ftr: IPHLPAPI was *forked* more than a decade ago from Wine.
Therefore no diff ever was guarded within those files whenever we implemented anything here.
The sum of X and width exceeds the X of next element.
So, change the width to <next x> - <current x>.
All affected translations were changed.
CORE-18829
- Use `HKCU\Software\Microsoft\Internet Explorer\Main\StatusBarOther` registry key for the setting persistence.
- Set its value to 1 when the status bar is visible, and to 0 when the status bar is hidden.
- Query it when the file browser is initialized to set the correct view state for the status bar.
- Set the default state of the status bar to hidden if the registry key does not exist, matching the behavior of Windows Server 2003.
CORE-19010
Store the state of the explorer toolbar, locked or unlocked, in the proper
registry location and set the proper toolbar state when initialized.
CORE-9094
- Set `HKEY_CURRENT_USER\Software\Microsoft\Internet Explorer\Toolbar\Locked`
to 1 when the explorer toolbar is locked, 0 when the toolbar is unlocked.
- Query `HKEY_CURRENT_USER\Software\Microsoft\Internet Explorer\Toolbar\Locked`
when the toolbar is initialized to set the correct locked state.
- Set the default state of the toolbar to locked if the registry key does not
exist, matching the behavior of Windows Server 2003.
In 0.4.14-dev-955-g 1cf564c25f
Katayama experimented with populating explorers file-menu when no object is selected.
Later we found out, that none of the new entries introduced by that commit really
made sense and even created duplicates. So the commit was reverted by
0.4.15-dev-6039-g 0fa4edebd9 'Revert CDefView::FillFileMenu (#5278)' CORE-18429
But it seems that not all parts were properly reverted back then,
maybe because 6 lines of new code were written between the two lines in the meantime.
The conime.exe program exists in WinXP/Win2k3 and it realizes Console IME. conime.exe calls kernel32!RegisterConsoleIME and kernel32!UnregisterConsoleIME. To realize Console IME, these two functions are required. CORE-11700
CORE-13951
MIDIMAP_LoadSettings and PlaySound:
Use RegOpenCurrentUser to access the appropriate HKEY_CURRENT_USER key.
This allows functions that use HKEY_CURRENT_USER to be called from system
processes while impersonating, for example: winlogon.
PlaySound:
- Open the correct registry key.
- Expand environment strings if registry value type is REG_EXPAND_SZ.
- Handle "SystemDefault" event and improve flags handling.
- Don't use proc_PlaySound as thread proc.
- Fix SND_ASYNC flag for system processes.
Move get_mmio* code out of proc_PlaySound to fix impersonation problem.
CORE-18906
- Increase the vertical height of the disk space pie chart and center
the "Drive %c" label below it.
- Add a 1 pixel outline at the sides and bottom of the disk properties
pie chart.
- Explicitly set the outline pen instead of relying on the DC's default pen.
- Widen "Disk Cleanup" button on ja-JP locale, to prevent the text from
being clipped due to the larger font size and spacing.
As on Windows, gives precedence to the user's extension in file names if there is one.
Otherwise appends the selected extension. If not selected, append default extension.
Correctly display the composition window and the candidate window.
- At update_caret function, set the position and font of the composition window.
- We don't use internal composition string. Rely on the composition window.
- Improve WM_IME_SETCONTEXT, WM_IME_CONTROL, WM_IME_SELECT, WM_IME_STARTCOMPOSITION, WM_IME_COMPOSITION and WM_IME_ENDCOMPOSITION message handlings.
CORE-11700
The changes of this PR are on EDIT controls.
- Delete composition_len and composition_start members.
- Add EDIT_ImmSetCompositionWindow helper function.
- At EDIT_SetCaretPos, set the position of the composition window.
- We don't use internal composition string. Rely on the composition window.
- Improve WM_IME_STARTCOMPOSITION, WM_IME_COMPOSITION and WM_IME_ENDCOMPOSITION message handling.
CORE-11700
Provide graceful handling of RegCreateKeyExW when lpSecurityAttributes->nLength is zero.
This reverts commit cc0716b65 (SVN r24491).
JIRA issue: CORE-15471
Improve compatibility.
- Rename window class "rundll32_window" as "RunDLL".
- Delete some shell32 codes about this window class.
CORE-13895, CORE-18350
- Call ImmSetCompositionFontW in WM_SETFONT handling of EDIT controls if necessary.
- If the specified font is NULL, then use DEFAULT_GUI_FONT.
CORE-11700
There is a bug, which displays empty optical drive as a valid Drive where applications can be installed, which is wrong. This change will skip displaying any drives that are not fixed in the disk usage table in the msi installer UI.
JIRA issue: CORE-18758
Proposed changes
Revert to old solution from Doug Lyons which was working fine.
40 is what en-US uses, and most other langs used that meanwhile as well. Some were forgotten to be synced though.
While comparing all the files, use the opportunity to slightly improve some headers also.
CORE-18382
Original patch is used to fix and improve french translation and layout.
The original (erroneous) horizontal shift of the second checkbox is kept.
It should be fixed homogeneously in all translations at once later.
- make all 3 input boxes IDC_COMPUTERNAME, IDC_ADMINPASSWORD1, IDC_ADMINPASSWORD2 on this page the same length for all languages and start and end at same x-coord, e.g. sk-SK.rc, uk-UA.rc
- let the input boxes end at the same x as the edits in the previous dlg IDD_OWNERPAGE
- respect the right margin of the dialog with all controls for all languages, makes it look more harmonic to previous and next dlg, see video:
- make the input fields as wide as possible, but not wider, depending on what the language allowed I used width of 139, 148, 163 (163 is same as the EDITs in IDD_OWNERPAGE used)
- fix too small x offset for some controls in some languages like bg-BG.rc, that looked very off
- in zh-CN.rc fix the maximum allowed length for computer name from 63 to 15
- ja-JP.rc 15 is max chars for computer name, not 63
- Fix it-IT text truncation "Organizzazione" in IDD_OWNERPAGE
- pl-PL.rc Steal some wisdom from 0.4.9 to better stick with en-US.rc dimensions.
syssetup/lang/lt-LT.rc is a 100% copy of en-US.rc. Not a single string or path within did differ.
The file emerged once during some refactoring when parts from userenv were moved to syssetup in SVN r73518 == git c24dcb3ff2 .
By deduplicating that file syssetup.dll shrinks for GCC8.4.0dbg RosBEWin2.2.2
and it will speed up the development of that module if one rc-file less has to be groomed everytime something is added into syssetup.dll.
It is quicker to recreate that file later if some use for translation would evolve in the module.
I linked this commit against CORE-18893 although my commit did neither introduce
nor would solve that bug of 'the language not generating any startmenu entries'!
I stumbled over the file being a total copy of en-US
after that ticket was created already and then compared the contents of lt-LT.rc to en-US.rc
So I though the linking would make kind of sense.
I installed in lithuanian before and after my commit. No change!
lt-LT.rc was the last one of such rc copy-pastes in [SYSSETUP], I double-checked them all now.
syssetup/lang/fi-FI.rc is a 100% copy of en-US.rc. Not a single string or path within did differ.
The file emerged once during some refactoring when parts from userenv were moved to syssetup in SVN r73518 == git c24dcb3ff2 .
By deduplicating that file syssetup.dll shrinks from 1.374.208 -> 1.361.920 for GCC8.4.0dbg RosBEWin2.2.2
and it will speed up the development of that module if one rc-file less has to be groomed everytime something is added into syssetup.dll.
It is quicker to recreate that file later if some use for translation would evolve in the module.
I linked this commit against CORE-18893 although my commit did neither introduce
nor would solve that bug of 'the language not generating any startmenu entries'!
I stumbled over the file being a total copy of en-US
after I saw that glitch and created the ticket and then compared the contents of fi-FI.rc to en-US.rc
So I though the linking would make kind of sense.
I installed in finnish before and after my commit. No change!
For: hivedef.inf, powercfg.cpl, sysdm.cpl
and: netshell.dll, netcfgx.dll, shell32.dll, userenv.dll
Also fix "My Documents" and "My Network Places" folder names.
* [SHELL32] Fix issue when creating a new file or folder. No chance to rename the displayed name.
The CNewMenu::SelectNewItem method starts with a request to notify the View object about a new item by calling SHChangeNotify.
After creating and obtaining the new PIDL by calling ILFindLastID, the CDefView::SelectItem method is called.
This method fails right from the start when calling LV_FindItemByPidl which returns -1, thus preventing the user from being given the opportunity to change the name displayed by default.
This failure is due to the fact that this object has not yet been added to the CDefView::m_ListView list. This is caused by the asynchronous nature of SHChangeNotify.
Thanks to Giannis Adamopoulos for the big help.
- Fix the visual glitch of the network settings dialog of 2nd setup.
- Shrink the IDC_NETWORK_TYPICAL and IDC_NETWORK_CUSTOM radio buttons of IDD_NETWORKSETTINGSPAGE resource dialog vertically not to be overlapped.
CORE-18874
de-DE: Harmonize the length of IDC_COMPUTERNAME with IDC_ADMINPASSWORD1 and IDC_ADMINPASSWORD2
which looks better and is already done like that in en-US.
de-DE: Also harmonize some 2nd stage captions. Some were named "Setup" and some were
named "Installation". I harmonized to "Setup" everywhere as that is what is also used in
1st stage, is shorter and no german person will have any issue understanding
that. It is also more likely to be kept in sync in the future if new dlgs would
be added, as it is the same word as used in en-US.
de-DE: Align the IDC_PROJECTS with the GPL button (which is slightly more wide in de-DE).
All languages:
Use same width for IDC_PRODUCT_OPTIONS as for IDC_PRODUCT_DESCRIPTION.
Win2K3 just wraps in SEH only this syscall, which appears pointless
and was removed since NT6.0. We remove it too as an experiment.
Revert the change that added SEH here from commit 6d97180 (r52899).
It regressed in master by 0.4.15-dev-5613-g 7a17c7d9ad
It was the only place within the ros sources where message WM_SHOWSTATUSDLG
was sent also from code. And therefore the only one that needed to be updated as well.
Based on earlier PR #4704
- Fixed alignment for too long text.
- Improved translation.
- Minor adjustments.
CORE-18381
Co-authored-by: Hermès BÉLUSCA - MAÏTO <hermes.belusca-maito@reactos.org>
Co-authored-by: Kyle Katarn <contact@kcsoftwares.com>
This PR will reduce the failures of CommandLineToArgvW testcase of shell32_apitest.
- Use isspace and isblank in shell32!CommandLineToArgvW.
- Strengthen shell32_apitest:CommandLineToArgvW testcase.
CORE-17787
When using wine internet explorer it crashes if you click print even if you click cancel, this fixes that issue.
nsIWebBrowserPrint_Print apparently causes issues for unknown reasons.
JIRA issue: CORE-16884
By returning S_OK print doesn't cause a crash.
Use ERR and S_OK return before hitting nsIWebBrowserPrint_Print
German accelerator for "&Datei" collided with "A&dresse".
Fix by switching to what MS uses in german XPSP3.
see the before-state with the [BROWSEUI] collision here:
https://jira.reactos.org/secure/attachment/65116/ros.png
Please note that in current master head the accelerators in
[BROWSEUI] filebrowser are generally broken for other reasons.
So this commit will not entirely fix CORE-18824
Also update to new-style header (inspired by es-ES.rc) and
https://reactos.org/wiki/Coding_Style
and add a missing translation.
When no extension is provided, add default extension with correct precedence using PathFileExistsDefExtW.
- Use PathFindOnPathExW when searching in current directory and PATH environment variable.
- Fix Width String terminator to UNICODE_NULL.
- Fix all failed tests on PathFindPathExW KVM test.
CORE-17612
[BROWSEUI] German accelerator for "F&avoriten" collided with "&Ansicht".
[EXPLORER] German accelerator for "F&avoriten" must be "&Favoriten" also in the startmenu.
[BROWSEUI] German accelerator for "S&chnellstartleiste" was missing.
Fix all 3 by switching to what MS uses in german XPSP3.
see the before-state with the [BROWSEUI] collision here:
https://jira.reactos.org/secure/attachment/65116/ros.png
Please note that in current master head the accelerators in
[BROWSEUI] filebrowser are generally broken for other reasons.
So this commit will not entirely fix CORE-18824 yet, but just
rules out the small related glitches in the german rc-files.
- Add uFlags argument to CDefaultContextMenu::AddShellExtensionsToMenu- CCopyToMenu, CMoveToMenu, and CSendToMenu check the uFlags against (CMF_NOVERBS | CMF_VERBSONLY).
CORE-16544
The code was trying to check whether the output string was already NULL terminated by RtlUnicodeToMultiByteN before NULL terminating it by checking DataStr[*count - 1] for a NULL terminator. But since RtlUnicodeToMultiByteSize always returns the size without the NULL terminator, DataStr[*count - 1] would always be the last actual character, never an optional NULL terminator.
For 0 sized strings this would actually lead to accessing the output buffer at position -1 (on 32 bit) or 0xFFFFFFFF (on 64 bit).
Fix this by removing the check. This fixes a crash in advapi32_winetest:registry on x64.
Use the same width for COMBOBOX IDC_TIMEZONELIST that we use in timedate.cpl to
prevent strings from getting cut off for some timezones with long names.
To still make it look appealing afterwards, enlarge and move a bit to the right also IDC_TIMEPICKER.
IDC_AUTODAYLIGHT used different widths in several translations, we can harmonize
for all languages to the biggest one that was used: 230 (taken from eu-ES.rc).
And also fix a few additional whitespace-glitches in zh-CN, zh-HK, zh-TW and en-GB.
After this commit the whole dialog IDD_DATETIMEPAGE uses the exact same dimension for
all of its controls for all languages.
It is good practice to not make the expanded combobox larger than the dialog itself.
CBS_NOINTEGRALHEIGHT will allow to set a fixed amount of pixels for the v6 combobox,
and therefore allows us to hide the problem for this specific dialog.
The problem was unhidden by 0.4.12-dev-882-g e3e173ffaa
Even after this patch I think Wines comctl32 can be improved for comboboxes with very
many elements that do not use this small trick yet. I will create a follow-up-ticket.
- Add enums for Control Panel and Registry Folder columns
- Fix iColumn values in GetDetailsOf() and fix checks in
GetDefaultColumnState()
Current CControlPanelFolder::CompareIDs was using wrong lparam column
index. Actual column index should be 1 instead of 4. Because of this
the comment column next to name in details view was not being displayed
correctly. The same fixes for CRegFolder class functions.
PR #4944. CORE-18743 CORE-18501
The old string lacked a *mandatory* comma before the word "übereinstimmen"
as the part "die Sie nutzen wollen" is a relative clause,
relative to the noun "Anwendungen".
But we can fix it without having to add the comma by some shuffling.
That also improves the reading flow.
Fix a bug when after closing the settings window for a screensaver,
the preview for it wouldn't restart.
- Keep waiting for messages from the settings window while it's open
- Fix leaking two handles to avoid creating zombie processes
CORE-18680
Several fixes and improvements to the CDeskLinkDropHandler:
- Set default working directory for shortcuts (except folders and zip files)
- Copy existing shortcut to the desktop if the source file is a shortcut
- Prevent destination file name collision
Verified on Windows XP SP3 and Windows 7 SP1.
[NEWDEV] Enable OK button in "Browse For Folder" only when driver is found
Implement BrowseCallbackProc() function which sends BFFM_ENABLEOK message
to the browse dialog whether the driver is found in the selected folder.
Pass the search path to the browse dialog depending on the current index
of the drop down combobox. If the index is not set, just get window text.
Then, automatically expand the tree view to the specified path by sending
BFFM_SETSELECTION message.
Also fix a bug in SearchDriverRecursive() where a duplicate backslash
was added to the PathWithPattern string variable.
[SHELL32] Do not add Recycle Bin to the tree view items in FillTreeView()
[SYSSETUP] Add source path to the "Installation Sources" multi-string key
Each time the ProcessSetupInf() is being called, add the source path
to the "Installation Sources" registry key, if it's not added there yet.
The driver search path combobox will be then populated using its value.
xml2: Import upstream release 2.10.0.
wine commit id 015491ab32742ace5218d37b1149c58803858214 by Alexandre Julliard <julliard@winehq.org>
Note: Upstream msxml3_test has switched away from WINE_NO_LONG_TYPES, so
I've kept the old printf format specifiers for now. Once we do a full
sync, we can get rid of __ROS_LONG64__ for this test and use them
unmodified.
Makefile.am: this hasn't been updated in a while
security.c: WIN32 -> _WIN32 to keep the ROS-diff consistent with the rest
win32config.h/libxslt.h: remove unnecessary ROS-diff
xsltwin32config.h: this was missed in the 1.1.34 sync
xsltexports.h: mark a ROS-diff as such
CDefaultContextMenu::DoProperties provides a fallback call
to the property sheet testing the return value of the _DoCallback method,
which is ultimately the return value of SH_ShowDriveProperties().
SH_ShowDriveProperties() sometimes returns an HRESULT, however it is marked
as returning a BOOL. Then, DrivesContextMenuCallback() always handles this
result as an HRESULT.
Fix SH_ShowDriveProperties() to always return a BOOL as it is intended,
and in DrivesContextMenuCallback() handle the result accordingly.
CORE-18537
This fixes the toolbar in FileZilla 3.8 being drawn wrong (grey)
It regressed by 0.4.15-dev-1603-g 232c45fcd7
And todays fix is a partial revert of that guilty rev.
The reverted part is not even necessarily needed for what we had in mind back then,
It was just a first tiny step with the aim to get rid of the comctl32.h changes of that commit.
But that goal was and even is far out of reach for many other reasons also.
Actually we should have reverted the toolbar.c change back then already,
before committing the "flipfix9".
CORE-18263
[SYSDM] Prevent my german-teacher from getting a heart-attack by that misplaced comma.
[SMSS] Prevent myself from getting a heart-attack by that superfluous dot.
Furthermore this addendum serves the purpose of actually linking both PRs
and their 4 previous commits to the actual JIRA ticket CORE-18754, which wasn't
the case before:
0.4.15-dev-5392-g 04b2d35f5b
0.4.15-dev-5391-g a8e06d92e8
0.4.15-dev-5390-g a4274ad548
0.4.15-dev-5389-g 5dc43c0f32
- Implement the architecture-specific pagefile size limits.
In particular, verify that the selected volume on which to create
the page file can accomodate the theoretical maximum limit (e.g.
FAT32 cannot accomodate a 16 TB pagefile on x64, and the limit must
be lowered down to 4 GB).
- Change the IDS_WARNINITIALRANGE and IDS_WARNMAXIMUMRANGE strings
so that the maximum limit mentioned is the dynamic one.
- Review, improve and fix other aspects of the code.
Parts of this PR: addendum to commits 4d2d2dbb2 (#2597), 3bee3b92a (#2706).
DFM_MERGECONTEXTMENU handler works better now. However there is difference
between ours and Windows' menu building systems, which has to be fixed.
Addendum to 64657051c3. CORE-13841 CORE-18577
In explorer filebrowser 'details' view
the column 'comments' had invalid length of zero, and therefore
was hidden under the 'attributes' column.
I guess even after this patch, it will not display any
sane contents in there yet.
Therefore I chose a small width of 10 only, to reduce the chance for
it to trigger an undesired horizontal scrollbar.
As far as I see we don't support manual showing and hiding of additional columns yet.
By default my Windows does not show this column. But when activated manually, it is always
shown on the right hand side of the 'attributes' column.
Sync the barebone fix from Wine (wine-7.19-557-g13cc08e32d6):
https://source.winehq.org/git/wine.git/?a=search&h=HEAD&st=grep&s=ddraw_surface_update_frontbuffer%28
and
13cc08e32d:/dlls/ddraw/surface.c
- Added "swap_interval" argument to ddraw_surface_update_frontbuffer()
- Added ddraw_swap_interval_from_flags() function for dispatching
the integer type of required swap interval
- Updated ddraw_surface_update_frontbuffer() function for managing the
swap interval between backbuffer and frontbuffer (and relative flipping)
- Updated the ddraw_surface7_Flip() function for relaying the swap interval
to ddraw_surface_update_frontbuffer() function.
- Added 0 value (as Wine) when the swap interval is not required
This fixes empty white screen issue on rendering because of lacking
swap interval for the software / games that use ddraw.
CORE-18547
This DLL was exporting legacy NT-incompatible or ROS-specific SM client
functions, that have been since 10 years now (2012) replaced by the new
NT-compatible SM:
- SmConnectApiPort(): was just SmConnectToSm().
- SmCompleteSession():
The legacy SMSS used it for when a subsystem initialization was finished.
Now (NT-compatible) this function is called by subsystems **only** when a
subsystem session **terminates**: SmSessionComplete().
- SmExecuteProgram(): was just the client side of SmLoadDeferedSubSystem()
(whose server side is not implemented yet). The legacy SM "old" SmExecPgm
implementation actually was "SmLoadDeferedSubSystem"...
- SmLookupSubsystem(): is a utility-only function to read any registry value
inside "Session Manager\SubSystems".
Move SMDLL's readme into SMLIB and update its contents.
Collect some residual useful functions into smutils.c (and moved in SMLIB,
though not compiled yet):
- SmExecuteProgram(), now implemented as a wrapper around SmExecPgm();
- SmLookupSubsystem(), described above;
- SmQueryInformation(), that retrieves a list of currently-running subsystems.
[SMLIB] Validate SbApiPortName's length in SmConnectToSm().
Fix CommandLine length validation in SmStartCsr().
Add documentation (+ SAL annotations) to the NT-compatible SMSS client functions.
smmsg.h: Add both Win32 and Win64 struct sizes C_ASSERTs for those whose size
change between these two processor architecture sizes.
[SMLIB] Introduce SmSendMsgToSm() as helper to send data into the SM LPC port.
+ Make the other API functions use it.
It should be observed that in Vista+, both functions SmConnectToSm() and this
new SmSendMsgToSm() are exported by NTDLL under the names RtlConnectToSm()
and RtlSendMsgToSm() (and use the same signature).
See: https://www.geoffchappell.com/studies/windows/win32/ntdll/history/names60.htm
[NTDLL] Correctly stub RtlConnectToSm() and RtlSendMsgToSm().
[NTDLL_VISTA] Link to SMLIB and simply export RtlConnectToSm() and RtlSendMsgToSm().
Partially revert some aspects of commits 5696e4ba4 and bf40c7a31.
(See PR #4340.)
In order for Win2k3 kernel32.dll to operate with our basesrv.dll (or our
kernel32.dll to operate with Win2k3 basesrv.dll), we need in particular
to have the CreateNlsSecurityDescriptor() helper to exactly take the
expected parameters. Namely, a pointer to a **user-allocated**
SECURITY_DESCRIPTOR buffer, its size (and an access mask).
The function expects its caller to provide all this, and the caller expects
the function to initialize the security descriptor buffer. Note that the
function does *NOT* allocate a new descriptor buffer to be returned!
Indeed, with the way it currently is in master, using Win2k3 kernel32
with our basesrv is now failing with the errors:
```
NLSAPI: Could NOT Create ACL - c0000023.
(subsystems/win/basesrv/nls.c:279) NLS: CreateNlsSecurityDescriptor FAILED!: c0000023
NLSAPI: Could NOT initialize Server - c0000023.
(dll/ntdll/ldr/ldrinit.c:867) LDR: DLL_PROCESS_ATTACH for dll "kernel32.dll" (InitRoutine: 77E40D95) failed
```
(and, if we ever attempted to increase the so-claimed "dummy parameter"
descriptor size in the basesrv call, we would end up with its stack
corrupted and a crash).
Conversely, using our kernel32 with Win2k3 basesrv, would end up with
basesrv receiving a wrongly-initialized descriptor that would not work
(the buffer not being initialized with the contents of a descriptor, but
instead receiving some address to a descriptor allocated somewhere else).
