- CDefViewBckgrndMenu : Add class to handle the background menu of the CDefView and handle stuff that previously were handled by CDefaultContextMenu. Let the shell folders fill their own part of the background context menu.
- CDefView: Create a CDefViewBckgrndMenu instead of a CDefaultContextMenu which was completely wrong.
- CDesktopFolder, CFSFolder: Implement returning the background context menu in CreateViewObject. Note that I am not completely sure that the callback should be adding the Properties item.
- CDefaultContextMenu: Remove code that was moved to CDefViewBckgrndMenu and try to improve the overzealous debug prints regarding the shell extensions.
CORE-12410, CORE-12472
svn path=/trunk/; revision=73405
- The new ScmDeleteServiceKey (r73400) and the already existing ScmDeleteRegKey are the same functions: remove the old ScmDeleteRegKey and use ScmDeleteServiceKey instead AND rename ScmDeleteServiceKey to ScmDeleteRegKey (as it can be used generically within services.exe).
- Call RegDeleteKeyW for the subkey after we have closed its opened handle.
svn path=/trunk/; revision=73401
Partial implementation of CMP_RegisterNotification and CMP_UnregisterNotification.
[UMPNPMGR]
Partial implementation of PNP_RegisterNotification and PNP_UnregisterNotification.
CORE-12217 #comment This should fix one of the issues.
svn path=/trunk/; revision=73394
A hack has gained enough XP and levels up! Congratulations to the MmCreateSection() in charge of forcing file cache initialization.
As show by kmtest:NtCreateSection tests and real world FSD (starting with MS FastFAT and Ext2Fsd), caching of a file can be initialized rather late (ie, on first effective read/write).
That means that our current hack is totally opless on newly created files with 0-size where the IRP_MJ_READ is most of the time opless too.
The hack, thanks to its level up, can now force a write (1-byte) in case read was unsuccessful and end of file.
In other words, this is the hack of a hack (and a hack v2).
It fixes a few failing kmtests.
Thomas, Peter, please retry the FSDs on which you are currently working on and report. Thanks :-)
CORE-11819
CORE-12475
svn path=/trunk/; revision=73384
- Automatically define __ROS_LONG64__ on LP64 systems. This makes no difference for ROS but makes things easier for downstream users of our headers.
Patch by Cat Zimmermann.
svn path=/trunk/; revision=73383
Tests show that the Print Provider functions are less forgiving and expect winspool.drv and spoolss.dll to perform the necessary sanity checks.
Also they don't default to the current environment if none was supplied as pEnvironment. Change our functions accordingly to pass the tests.
[LOCALSPL_APITEST]
Add a test to prove that spoolss.dll isn't just a forwarder, but also performs some sanity checks even though winspool.drv and localspl.dll also do.
This rather belongs to spoolss_apitest. But it needs to be tested inside the Spooler Server Process and it's not worth it to copy all the DLL injection code to spoolss_apitest just for a single test.
[SPOOLSS]
Check for an invalid buffer in GetPrintProcessorDirectoryW.
[WINSPOOL]
* Check the level before the RPC call in GetPrintProcessorDirectoryW.
* Supply our current environment if the caller didn't give any in EnumPrintProcessorsW and GetPrintProcessorDirectoryW.
* Implement GetPrintProcessorDirectoryA.
[WINSPOOL_APITEST]
Prove that pcbNeeded isn't touched when calling GetPrintProcessorDirectoryW with an invalid level.
ReactOS now passes all API tests for GetPrintProcessorDirectoryA and GetPrintProcessorDirectoryW. These tests also cover all known behaviors of both functions.
CORE-12399
svn path=/trunk/; revision=73373
Fix CONTAINING_RECORD calls for spoolss functions that map directly to localspl.
An innovative way to fail the RPC stack...
CORE-12449
svn path=/trunk/; revision=73371
In IRP_MJ_QUERY_INFORMATION/IRP_MJ_SET_INFORMATION, handle the case when we receive buggy FO without FCB.
This fixes bugchecks when using FileSpy.
CORE-12448
svn path=/trunk/; revision=73362
