git snapshots will contain the short version of the hash
that was used to generate the snapshot in their version number,
take that into account in the spec file.
Fedora has all the mingw dependencies for vdagent packaged but don't
want to ship mingw Windows binaries. This .spec can be used to run
local mock builds of vdagent using mingw, or scratch builds on koji, ...
read & write are async, and their completion is handled by handle_event(),
which returns status used by service execute loop.
previously an error in GetOverlappedResult caused vdservice hang.
rhbz#839564
User Interface Privilege Isolation is usually used only for specific windows of
system security applications (anti-viruses etc.), so with this patch mouse will
be irresponsive for these windows but keep working for the rest. A complete
solution might be switching to server mouse mode while the agent is still active.
This is only part of the message corruption solution.
The other part is fixing virtio-serial / spice-qemu-char throttling code.
-replace write_[lock/unlock/completion] calls with [new/enqueue]_message
-remove clipboard specific _out_msg_* class members
-remove ugly loop - while (a->_out_msg && a->write_clipboard());
-add _message_mutex for message queue
-fix pending_write race using _write_mutex
-TODO: enqueue large message without dividing it to chunks in advance
rhbz #846427
Fix build by default when using mingw (when more compiler are
supported we could do case by case)
Since we are statically building, strip resulting binary (we
also use lto, whole-program but that doesn't seem to change
much..)
Waiting for a Windows event will not last if it is already set.
For example, the client may send clipboard_release() messages
while we are not waiting in on_clipboard_request(), and this will
SetEvent(clipboard_event)
The following clipboard request will thus not wait for the data,
resulting in an empty clipboard & paste for the guest application.
We could say there is fundamentally a race as there is no obvious
way to know if a received message is related to the current request,
but by reseting the event before waiting for new events to come, we
at least clear the past events.
When resizing desktop to best match, the agent fails to update
it's current mode width/height and then incorrectly report
different total_width/total_height than real screen dimensions,
then scaling input incorrectly results in mouse cursor offset.
https://bugzilla.redhat.com/show_bug.cgi?id=823019
Apparently, it comes with a recent version of C library (msvcrt90?),
but I don't know if we can/should ship also a MS dll. Probably it used
to work with an inlined version with VS headers.
The build can be tweaked to use the -non-_s variant with -DOLDMSVCRT
CFLAGS.
The convention in vdagent codebase seems to put a space before
the opening parenthesis in initialization list, this commit fixes
2 inconsistencies related to that.
mingw has a _ftime_s prototype in its headers, but no corresponding
symbol available at link time. Workaround this issue for now by
#defining it to _ftime. This is untested on win64 where the workaround
may not be needed.
VDAgent::_mouse_x and VDAgent::_mouse_y are both signed but are
always assigned values from VDAgentMouseState which are unsigned.
gcc warns about this so change them to unsigned values.
If we use char * instead of const char *, mingw gets confused when
calling the constructor with a const char *: it tries to convert it
to a mutable string through a std::string conversion and outputs
a warning.
DWORD is an unsigned long, but it's assigned -1 in various places.
mingw warns when comparing a DWORD value against -1. This commit
casts the -1 to DWORD to avoid the warning.
