[FLTMGR][KMTEST]
Squash and push my local branch across to master as the patch is getting a bit large. This is still WIP and none of this code is run in ros yet, so don't fret if you see ugly/unfinished code or int3's dotted around.
[FLTMGR] Improve loading/reg of filters and start to implement client connections
- Implement handling of connections from clients
- Implement closing of client ports
- Add a basic message waiter queue using CSQ's (untested)
- Hand off messages for the comms object to be handled by the comms file
- Initialize the connection list
- Add a registry file which will contain lib functions for accessing filter service entries
- [KMTEST] Initial usermode support for testing FS mini-filters
- Add base routines to wrap the win32 'Filter' APis
- Add support routines to be used when testing FS filter drivers
- Move KmtCreateService to a private routine so it can be shared with KmtFltCreateService
- Completely untested at the mo, so likely contains bugs at this point
- Add support for adding altitude and flags registry entries for minifilters
- Allow minifilters to setup without requiring instance attach/detach callbacks
- Add tests for FltRegisterFilter and FltUnregisterFilter and start to add associated tests
This should fix some virtual machines and real hardware machine with empty floopy drive
not being able to boot ReactOS (stuck while initializing floppy.sys).
This fixes a regression introduced in r70746.
It could be generalized to other interrupts, floppy controllers not being reliable.
For more information: http://wiki.osdev.org/Floppy_Disk_Controller
CORE-7935
CORE-12908
CORE-13080
CORE-10495 CORE-13672
- Fix the behaviour of the EXIT command, where it set the last errorlevel
value ONLY when it was called with the /b switch and otherwise kept the
ambient one, instead of always using the value that the user specified
on the command-line.
- When a batch file has terminated, make the Batch() helper returning the
errorlevel value so that, when the batch file has been started with the
CALL command, CALL can in turn set the correct errorlevel value.
Verified with respect to Windows' cmd.exe.
CORE-13974
This should fix situations where (for example):
command_1 | command_2 && echo Succeeded
should *NOT* run "echo Succeeded" if any of the command_1 or command_2 has
failed.
This also makes the ExecutePipeline() function on par with the other
"ExecuteXXX()" helpers.
Problem diagnosed by Doug Lyons; patch inspired by contributor 'cagey45'.
To be reverted if it brings in too many regressions, even though I tried to do my best to address them.
Leaving the old driver in place, for now; if no regressions are reported regarding CDFS in 0.4.8, then old driver could be dropped for 0.4.9.
I use a (not yet used) dir-id value of "80" for this purpose (the value
"8" being already used for /bin/testdata/ and of course values from 9 up
to 50+ are also already used... >_>).
