Pierre Ossman [Wed, 24 Aug 2022 08:31:42 +0000 (10:31 +0200)]
Use BufferedOutStream in more streams
Avoid duplicating all the memory mangement, and instead use the
BufferedOutStream as a base clase for all out streams that need an
intermediate buffer.
This commit changes the color of the edges of the dot cursor in the Java
version of vncviewer from fully transparent to white, bringing it in
line with the C++ native version.
Pierre Ossman [Mon, 4 Jul 2022 12:42:55 +0000 (14:42 +0200)]
Store artifacts from Github actions
Might be useful for testing in some cases. Note that the Windows and
Linux binaries will depend on libraries from the build environment. So
the user will need to match those manually.
Pierre Ossman [Tue, 5 Jul 2022 07:19:07 +0000 (09:19 +0200)]
Fix Xvnc dependency list (again)
The previous commit was incorrect in that it overrides automake's
dependency list, rather than adding to it. So we need to make sure
things are duplicated between Xvnc_DEPENDENCIES and Xvnc_LDADD as
appropriate.
Use the same approach as the Xorg binary to keep things familiar.
Pierre Ossman [Mon, 4 Jul 2022 11:59:42 +0000 (13:59 +0200)]
Add missing dependencies for Xvnc
These aren't crucial for a clean build, but must be set for Xvnc to be
properly rebuilt if anything in the Xorg tree changes. automake doesn't
properly deduce these automatically from Xvnc_LDADD, unfortunately.
Pierre Ossman [Tue, 28 Jun 2022 12:49:40 +0000 (14:49 +0200)]
Safely discard large (extended) clipboard contents
Avoid having to buffer everything we want to discard, and instead do it
piece by piece. This is more efficient, and avoids hitting any limits on
the buffering.
Note that this is safe here because we already know we have all the
compressed data. It would not be safe for a general input stream.
Pierre Ossman [Tue, 28 Jun 2022 12:47:39 +0000 (14:47 +0200)]
Properly restore cork state when changed
These streams both need to change the corking state temporarily, but it
is important it is restored to the previous state or things might get
messed up.
For the zlib stream it would just leave things uncorked, which still
works but is less efficient.
But for the TLS stream it might make things very unresponsive as the
corking might be left on permanently, delaying packets indefinitely.
Martins Mozeiko [Tue, 7 Jun 2022 07:37:54 +0000 (00:37 -0700)]
Workaround for MFT cropping in H264 decoding
It seems MFT h264 decoder does not support frame cropping. Which
means that if frame width or height is not multiple of 16 then
decoded output can potentially be used with wrong offset. This
code adds explicit parsing of SPS to extract cropping information
to use, and will apply cropping if reported size differs from
expected.
Pierre Ossman [Sat, 4 Jun 2022 12:19:07 +0000 (14:19 +0200)]
Add target_link_directory() compat function
We need this function to deal with pkgconfig files properly, but
unfortunately it doesn't exist until CMake 3.13, and we need to support
CMake 3.10. So add a hacky compatibility function for older systems.
Pierre Ossman [Wed, 6 Apr 2022 13:45:36 +0000 (15:45 +0200)]
Ignore socket whilst processing data
FLTK has a lot of synchronous stuff (like dialogs), which mean that the
main loop might be run recursively in some cases. If there is data
available on our socket then CConn::socketEvent() will be called
constantly in a busy loop.
Avoid this by removing socket notifications whilst we are processing
data.
Pierre Ossman [Wed, 6 Apr 2022 13:38:45 +0000 (15:38 +0200)]
Enable stream sanity checks in debug builds
Enable this automatically for developers so we increase the chance of
these problems getting caught. There is a risk of overhead though so
keep them disabled for release builds.
Pierre Ossman [Wed, 6 Apr 2022 13:23:50 +0000 (15:23 +0200)]
Don't trust GNUTLS_E_AGAIN
Unfortunately this error can be given by GnuTLS even though the
underlying stream still has data available. So stop trusting this value
and keep track of the underlying stream explicitly.
Pierre Ossman [Fri, 11 Mar 2022 07:31:25 +0000 (08:31 +0100)]
Change vncsession startup error message
We need to get to the point of starting the session script before we
consider things a success. So this can fail in many different ways, not
just the daemonization. Adjust the error message to something more
generic to reflect this.
Zdenek Pytela [Mon, 7 Feb 2022 09:45:41 +0000 (10:45 +0100)]
SELinux: use /root/.vnc in file context specification
Instead of HOME_ROOT/.vnc, /root/.vnc should be used
for user root's home to specify default file context
as HOME_ROOT actually means base for home dirs (usually /home).
Pierre Ossman [Tue, 25 Jan 2022 08:56:31 +0000 (09:56 +0100)]
Improve logging around screen resizing
Avoid duplication, and add logging for some more error conditions so
it's possible to see in the logs why a resize has failed or produced
unexpected results.
William Sjöblom [Wed, 19 Jan 2022 09:04:53 +0000 (10:04 +0100)]
Fix session resize after mirroring on Linux vncviewer
If monitor mirroring was enabled while in a session with vncviewer
running on Linux, the session would not be properly resized on the
server. This was a consequence of only looking at the size and
coordinates of each screen when matching against existing screens after
the screen layout was changed, when in fact we have two (or more)
monitors with the same coordinates and size (but differing ids). This
led to the same monitor being added twice to the layout which would
later fail layout validation, resulting in no resize request being sent
to the server.
When matching, we now also check if the existing screen is not already
present in the layout before considering it a match.
Pierre Ossman [Tue, 18 Jan 2022 12:38:41 +0000 (13:38 +0100)]
Fix handling of VMware cursors
This is a regression from ad0f061. If a VMware cursor rect was split up
over multiple read()s then the stream would become corrupted as we set
the restore point at the wrong place.
Pierre Ossman [Wed, 12 Jan 2022 16:44:20 +0000 (17:44 +0100)]
Improve full-screen workaround on connect
If we've explicitly requested a position (using -geometry) then we were
still triggering legacy full-screen request logic. This is very counter
intuitive as having a position that doesn't align with a monitor should
disqualify us for that.
It turns out that at least marco (and probably everything
metacity-derived) ends up ignoring the requested position and we go back
to triggering the legacy logic again.
Work around this by assuming that the window manager will adjust our
initial position and don't consider that in the workaround logic.