By taking alternative code-paths in WdmAud and bypassing Sysaudio.
This relies on the patch 0.4.15-dev-765-g
b8e936a57b
It fixes the following bugs/regressions:
-CORE-13202 Unhandled exception from wdmaud.drv when recording sound in Scratch 1.4 leads to app-crash
-CORE-13488 A deadlock in "DiabloII" character selection screen and "The Lion King II"
-CORE-8726/CORE-9986/CORE-16564 AC97 driver is now working in the same session where driver was installed, no reboot-orgies needed anymore
-CORE-9981 almost 100% fixed, DosBox + Commander Keen6 properly plays music instead of garbled output,
same improvement for ScummVM 2.0 with Monkey Island 2
The playback is not yet *entirely* perfect, still a few hiccups now and then, but
by orders of magnitude better than before.
shell32.ShellMessageBoxW redirects to shlwapi.ShellMessageBoxWrapW.
However the latter one (shlwapi) is exported by ordinal only.
Trying to use in shell32.spec file the following syntax:
182 varargs ShellMessageBoxW() shlwapi.ShellMessageBoxWrapW
will therefore fail at runtime, because Windows/ReactOS will fail to
snap shlwapi.ShellMessageBoxWrapW.
Using instead an export by orginal:
182 varargs ShellMessageBoxW() shlwapi.#388
actually fails as well, but at link time, by both MSVC' LINK and GNU's
dlltool.
This generates a .def file containing the line:
ShellMessageBoxW=shlwapi.#388 @182
It is not clear why it fails, because according to:
https://docs.microsoft.com/en-us/cpp/build/reference/exports?view=vs-2019
and
https://devblogs.microsoft.com/oldnewthing/20121116-00/?p=6073
this should be possible.
However one encounters the same errors as those described at:
https://groups.google.com/forum/#!topic/microsoft.public.vc.language/AVnx-EnYdsY
and
https://www.xspdf.com/questions/22333.shtml
...
So instead we use another trick, that appears to be already in use in
the shell32.spec. Since the shlwapi functions are imported because we
link to it, this means that ShellMessageBoxWrapW() is already available
through this import. So we can specify it in the .spec file as is,
without a full name prefixed with "shlwapi." .
Therefore the associated shell32.ShellMessageBoxW function will instead
link to the small ShellMessageBoxWrapW() import stub that is generated
automatically.
They are currently specified for documentation purposes (i.e. what
Windows 8+ CMD.EXE can report) but not used yet, since ReactOS does not
support them.
This makes ShellMessageBoxW use the correct implementation where the
text buffer size is dynamic, instead of having a too small hardcoded
size.
Fixes CORE-17271.
See also PR #3172 by Kyle Katarn, supplemented with some ideas from
Mark Jansen.
However we cannot straightforwardly implement ShellMessageBoxA around
ShellMessageBoxW, by converting some parameters from ANSI to UNICODE,
because there may be some variadic ANSI strings, associated with '%s'
printf-like formatters inside the format string, that would also need
to be converted; however there is no way for us to find these and perform
the conversion ourselves.
Therefore, we re-implement ShellMessageBoxA by doing a copy-paste ANSI
adaptation of the shlwapi.ShellMessageBoxWrapW function.
Note that, on Vista+ onwards, shlwapi implements both ShellMessageBoxA/W,
and shell32 directly forwards these exports to shlwapi, thus avoiding
these workarounds.
[PSDK] Explicily use WINAPIV for the variadic ShellMessageBoxA/W functions.
[INCLUDE/REACTOS] Add ShellMessageBoxWrapW in shlwapi_undoc.h .
Remove misuse of multithreading and use NtDeviceIoContol with completion apc
callback instead (mirrors usage of WriteFileEx in WdmAudCommitWaveBufferByLegacy).
This fixes a crash caused by race condition. Code was simulating completion
callback using a thread, this resulted in single threaded code being executed
simultaneously by multiple threads.
CORE-17214
* Move RtlRunOnce functions from kernel32_vista to rtl_vista and export them from ntdll_vista
* Move condvar.c and srw.c from ntdll_vista to rtl_vista
* Move ntdll_vista build script to a subfolder of ntdll
The RtlRunOnce functions are taken from wine, completely unmodified.
The code that was in kernel32_vista had change that used a global keyed_event handle, but was never initialized, so we were still passing NULL thus using the global ExpCritSecOutOfMemoryEvent.
The sorting in explorer broke by
SVN r73128 == git
24fcf531e7
Thanks to Doug Lyons for providing this patch.
I do intend to port it back into 0.4.14RC.
