The previous implementation passed `length - str_len` to `snprintf()`,
yet compared the return value to `length`. This is incorrect, as
`length` is not the buffer size provided to `snprintf()`.
When using the fr-be-azerty remote keyboard layout on an RDP connection,
the dash ('-') and underscore ('_') are swapped.
Underscore and dash are located on the same key on a Belgian azerty
layout. Dash should be the normal/unshifted character, and underscore
should be the shifted character. The current mapping has this the other
way around, so let's fix this.
Signed-off-by: Sander Vanheule <Sander.Vanheule@UGent.be>
Older versions of libvncclient did not free all memory within
rfbClientCleanup(), but this has been corrected as of their 0.9.12
release. As guacamole-server may well be built against older versions of
libvncclient, we can't simply remove the manual free() calls, but we
should be sure to set any memory that we free ourselves to NULL so that
rfbClientCleanup() does not attempt to free it again.
The changes introduced by GUACAMOLE-1181 (commit 2c86e20) were made
conditional as older versions of FreeRDP will automatically free the
wStream, resulting in a double-free if we attempt to do so ourselves.
The macro controlling that conditional code is defined within config.h,
which is missing here. Without that macro, the call to Stream_Free()
always occurs, and we get a double-free with older FreeRDP.
The libuuid library is widely available (part of util-linux) and much
more frequently updated. The OSSP UUID library works great, but was last
updated in 2008 and causes some confusion for users that have libuuid.
Debian stable currently provides a FreeRDP package that is a snapshot of
commit 2693389a103394e035edc2a01055ca2c9ccccb21, whereas
stable-backports provides a FreeRDP package from release 2.2.0. Only a
release or release candidate can be relied upon for consistent behavior.
It is expected under normal circumstances that the connection may be
abruptly closed, including during the handshake. For example, this will
commonly occur when a TCP load balancer is performing a simple service
health check.
An message noting that the connection has been closed during the
Guacamole protocol handshake is really only of benefit when debugging,
where that information may provide useful context. If not debugging, the
message amounts to log noise.
There is a small typo in src/libguac/tests/parser/read.c, src/libguac/tests/socket/fd_send_instruction.c, src/libguac/tests/socket/nested_send_instruction.c.
Should read `verifying` rather than `verfying`.
Fixes https://issues.apache.org/jira/browse/GUACAMOLE-1225
