function. Because synergy doesn't use posix cancellation, it
cannot wake up a thread waiting on a condition variable. So
the wrapper would wake up periodically to test if the thread was
cancelled (according to synergy's cancellation state) then go
back to waiting. Well the condition could be signalled while
we're testing and be lost and we'd never return from the wait.
So now we wake up after a maximum timeout and return to the
caller. The caller must check for this spurious wakeup but all
callers should do this anyway so we're okay there.
an strange deadlock seen on OSX. The CSocketMultiplexer deadlocks
with two threads, one waiting for m_polling to become false and
the other waiting for m_pollable to become true. The weird part
is that they're both false so the first thread should proceed.
It either didn't receive the broadcast when m_polling went to
false or it's not really checking the actual value of that flag.
I can't see how the former is possible and this change fixes the
latter.
modifiers for all keys for testing purposes and forgotton to fix
that. Now choosing required modifiers per key. This fixes
shift+arrow keys suppressing the shift key and, i think, the
option key not working.
the handling of compose key sequences. The key presses were
suppressed but not the corresponding releases, confusing the
clients. It also adds support for generating keysyms via the
compose key if the necessary dead keys or Mode_switch are not
available.
when leaving screen. This may help determine how to avoid
DOS command prompt windows being in the foreground when leaving
the screen since they suppress handling of the shift key.
fix some other bugs that involve unexpectedly quiting. The problem
was that synergy would (cleanly) quit when receiving an event with
a message id of 0 when not running as a service.
the opposite handedness if the desired handedness is missing.
For example, if Alt_R is unmapped as it often is on keyboards
with Mode_switch, the client will use Alt_L if it is mapped
when told to synthesize Alt_R. This should fix handling of
AltGr sent from a win32 server. AltGr is literally just
Control_L and Alt_R and, previously, an X11 client without
Alt_R mapped would only simulate Control_L. This does not
address the fact that the user may really just want the Alt
modifier; there are already hueristics to attempt to figure
that out.
and Xinerama correctly. Also was using '#if defined(...)' instead
of '#if ...' for testing configure macros in some places. This
yields the wrong answer if the macro is set to 0, which means
missing/disabled.
events when receiving any event with message == 0 when the system
doesn't use old style mouse wheel events. Some programs (especially
the flash plugin) would send events with message == 0 causing
undesired wheel scrolling.
the cursor wasn't being parked in the center of the main screen
but instead at the center of the total display surface. This could
place it off or dangerously close to the edge of the transparent
window that covers the main screen and prevent synergy from capturing
mouse motion.
duplex keys but broke primary screens. This fixes both and
also ensures that the primary screen updates its shadow toggle
modifier state when leaving so the secondary screens get the
correct toggle modifier state. Together these fix some strange
inconsistencies in toggle state across screens.
for class visibility on pointers to member function 'using the
qualifying scope of the name itself.' what this means is if method
'prot' is declared protected in class A and B inherits from A then
a method in B cannot use &A::prot but can use &B::prot. Synergy
now does this in the one place it had not.
keys are mapped to the same button. For example, the backslash
virtual key shares a button with some other virtual key on british
english key mappings. Synergy could end up using the wrong virtual
key. In the given case, the other virtual key produced no character
at all. To determine which virtual key should really be mapped to
a button we map the button back to a virtual key and see if it's the
virtual key we started with.
Also fixed mapping of pause key. Previously, windows+pause sent to
a win32 client wouldn't bring up system properties like it should.
non-blocking connect is available for writing (i.e. the connection
was successful) when the connection has actually failed. This
caused synergy to rapidly retry connecting. This change makes
synergy check for a connection error whether one was reported or
not. Thankfully, OS X does correctly set the socket error state
when it bogusly reports a successful connection so we can tell the
connection failed.
crs