The first part of raw transfer sequence is to transfer the format
list of the session A into the session B. Then we will be able to
request/reply with raw data using proper format IDs.
xf_cliprdr_server_format_list() of the session A now exposes the
raw server format list. As soon as the list is received, it is
serialized and put into _FREERDP_CLIPRDR_FORMATS property.
xf_cliprdr_get_requested_targets() of the session B now checks
whether the clipboard owner is a FreeRDP session with enabled
raw transfer capability. If it is, the raw format list is simply
extracted from _FREERDP_CLIPRDR_FORMATS of the clipboard owner.
Otherwise, the format list is populated from the usual TARGETS
clipboard format.
Some time ago there was a property _FREERDP_CLIPRDR_ID which was indended
to indicate that an XFreeRDP window owns a clipboard. This was necessary
for raw transfers. This property was used by xf_cliprdr_is_self_owned()
function. However, raw transfer support was broken and the meaning of
xf_cliprdr_is_self_owned() gradually changed into checking whether
the *current* window owns the clipboard, not just any XFreeRDP window.
Thus _FREERDP_CLIPRDR_ID was removed in a4580923e7 (xfreerdp/clipr:
fix self owned test and hardening).
However, now we are going to fix raw transfers and we need that property.
This patch reintroduces a similar property "_FREERDP_CLIPRDR_RAW" which
indicates that a window is an XFreeRDP window with enabled raw transfer.
It is currently used by xf_cliprdr_server_format_data_request() to
correctly request format data from another XFreeRDP instance via raw
transfer protocol.
This property can be queried from the clipboard owner with the function
xf_cliprdr_is_raw_transfer_available() and can be enabled or disabled
on the current window by xf_cliprdr_set_raw_transfer_enabled().
Disabling raw transfers will be necesary to correctly implement file
transfers in the future. However, currently raw transfers are always
enabled.
xf_cliprdr_send_data_response() is consistently called with NULL
data pointer as a way to report errors, but it was not setting
the msgFlags field accordingly.
Sometimes Windows sends strings with excess null terminating bytes.
For example, when one copies digits from calc.exe. At the same time,
some local applications freak out when they encounter null bytes
(at least LibreOffice is known to be replacing them with '#').
According to the specification of UTF8_STRING format [1], the string
data must not contain any trailing null bytes. So they all should be
trimmed, not only the last one.
Also, if the trailing null byte is not present, the length should not
be adjusted. For example, Firefox is actually sending "HTML Format"
without a null byte while Internet Explorer adds one. The spec for
text/html format [2] says nothing about the teminating null byte, so
we are free to remove it, but at least we should not mistakingly
delete '>' character of "</html>" tag when it is the last character.
[1] http://www.pps.univ-paris-diderot.fr/~jch/software/UTF8_STRING/UTF8_STRING.text
[2] https://www.ietf.org/rfc/rfc2854.txt
Clipboard formats are identified by numerical IDs and literal names.
We can keep using arbitrary defined IDs for local clipboard formats
as we are sure that they have some fixed meaning, but the server can
and will be using its own IDs, which can be different from ours for
the named formats.
Therefore:
1) A correct way to compare a local format to a remote one is
to check the names first, and only then compare their IDs.
(Extra care should be taken to support short format names.)
2) Server IDs cannot be used with wClipboard directly when dealing
with named formats. Format name should be used to extract correct
local ID for the use with Clipboard{Set,Get}Data().
Also, I find the notion of 'alternate' format IDs to be confusing.
We either deal with a fixed ID format (declared in <winpr/user.h>),
or a format that was given an arbitrary fixed ID for local use
(defined in <freerdp/channels/cliprdr.h>), or a remote format
identified by a pair of an ID and a name. Format IDs can be local
and remote, but there are no 'alternates'.
So now:
1) A new function xf_cliprdr_formats_equal() is used to compare
formats correctly in xf_cliprdr_get_server_format_by_atom()
when searching for a server format corresponding to a local
one, and in xf_cliprdr_server_format_list() when constructing
a local TARGETS list from the server format list.
2) Correct local format IDs are used with wClipboard conversions
by xf_cliprdr_process_requested_data() and
xf_cliprdr_server_format_data_response().
3) We refer to formatId and formatName when doing requests,
and srcFormatId and dstFormatId when doing conversions,
instead of using formatId and altFormatId for both purposes.
4) Server format ID and name are used to identify cached clipboard
contents. The name is compared directly as a pointer because it
will be a pointer from the same clipboard->serverFormats array.
Also, the clipboard contents are invalidated when format list
arrives, so xf_cliprdr_server_format_list() now also clears
the format ID and name together with the data.
The functions now have appropriate names which tell what exactly
they are searching for:
xf_cliprdr_get_client_format_by_id()
Get a client-provided format by client-side ID.
xf_cliprdr_get_client_format_by_atom()
Get a client-provided format by client-side format name.
xf_cliprdr_get_server_format_by_atom()
Get a corresponding server format by client-side format name.
The return types of functions have been adjusted accordingly and
correct formats are now used everywhere without mixing them up:
client-side formats are used for client -> server data flow,
while server-side ones are used for server -> client tranfers.
This resolves the issue #1414 as, for some reason, xfreerdp required
server format list to be present to be able to provide its own client
formats. Actually, we need only client format list to provide these.
Also, CF_RAW special case is handled in a more elegant way: it is
assumed to be present in every server format list (which is true).
This patch is needed when wanting to install binaries/libraries to other
locations than data. The linux distro Exherbo installs binaries and
libraries to /usr/<chost>/bin, /usr/<chost>/lib respectively but
manpages should still go in /usr/share/man/ because they are
architecture independent, without this patch they go in
/usr/<chost>/share/man unconditionally.
cmake documentation states:
DATAROOTDIR - read-only architecture-independent data root (share)
So this patch makes it use that so that its configurable.
- fixed wrong calculation of xfc->fullscreenMonitors.[right|bottom]
- only use _NET_WM_FULLSCREEN_MONITORS if at least 2 monitors are involved
- call XMoveWindow before setting the _NET_WM_STATE_FULLSCREEN property
- xf_cliprdr_is_self_owned() lied if multiple xfreerdp instances were
running.
- fixed a few unchecked callocs
- added/modified and handled some return values in compliance with
the new hardened channel api
1. Remove all uses of "localWindowOffsetCorr" variables, they added an extra layer of complexity and they are not actually needed to handle coordination of window position/size between
the local coordinate system and the remote one. This logic was causing issues in the case where the window was moved off the left side of the screen.
2. Update the xf_setWindowVisibilityRects function to offset the visibility rects as necessary when the window is hanging off the left side of the screen.
3. Stop sending mouse events when doing keyboard moves/sizes(as desired), and stop sending two mouse events for non-keyboard moves/sizes
4. Move location of new UTF8_STRING variable from previous commit
5. Refresh window and window shape for any window position/size updates, this helps keep the local and server windows in sync and works around some race conditions
of existing after only looking at part of the information. For instance, window visibility
rects are part of the message with the new order and were being ignored.
The regions used to store and calculate the invalidRegion are exclusive
of the bottom and right edges, not inclusive.
Fixes "mouse droppings" in mspaint.exe when moving the mouse leftwards
across the canvas.
remove duplicate call to XStoreName when setting window title
expand WITH_XEXT #define for rail window rects as extra unecessary work was being done when WITH_XEXT was not defined
1. Make use of freerdp_set_last_error to set authentication failure without the helper functions
2. Rename ssl callback function
3. Break out AuthenticationOnly exit handling from bad connect handling
Don't abort the entire xf_rail_window_common function when the window is
already in the correct location.
To reproduce:
- move an application off the edge of the screen
- resize the window to cause a shape to be set
- move the application window fully on the screen
- resize the application window larger
- note lack of drawing in newly enlarged portion of window
Bug introduced in abf6d4f71e "xfreerdp:
prepare RAIL migration away from libfreerdp-rail" when
xf_rail_MoveWindow was copy-and-pasted into xf_rail_window_common
without noticing that the "return" would omit the rest of the combined
function, not just the portion that was pasted.
Since REGION16 uses unsigned values, when appWindow->x or appWindow->y
is negative, the region will have a very large left or top value.
Avoid this problem by clamping to 0 before casting to an unsigned value.
When connecting to windows 8.1 machines the remote RDP server
implementation sometimes sends invalid H264 data. To avoid client
disconnections ignore the broken updates.
Channels like EGFX need resources like the main window in order to
work correctly. Before the window, GDI,.. is freed it needs to be
ensured that all channels are stopped properly to prevent them to access
already freed resources. Disconnecting the channels first fixed a
possible race condition/SEGFAULT that could occur with remote initiated
disconnects.
winsock.h pulls in a lot of defines and dependencies that are not
required and partially unwanted in winpr's core (for parts that are not
related to network). In order to get rid of this dependency and have an
independent defines for extended winpr functions the WINPR_FD_* defines
are used internally (and for exposed functions). Where required, like in
WSAEventSelect, the FD_* is mapped to WINPR_FD_*.
Passing True to XSync() discards any pending X11 events. Occasionally
this caused ButtonRelease or KeyRelease to be lost and not forwarded
to the remote computed, leading to stuck keys and buttons.
This should resolve issue #2391
This patch contains:
* checks for malloc return value + treat callers;
* modified malloc() + ZeroMemory() to calloc();
* misc fixes of micro errors seen during the code audit:
** some invalid checks in gcc.c, also there were some possible
integer overflow. This is interesting because at the end the data are parsed
and freed directly, so it's a vulnerability in some kind of dead code (at least
useless);
** fixed usage of GetComputerNameExA with just one call, when 2 were used
in misc places. According to MSDN GetComputerNameA() is supposed to return
an error when called with NULL;
** there were a bug in the command line parsing of shadow;
** in freerdp_dynamic_channel_collection_add() the size of array was multiplied
by 4 instead of 2 on resize
gcc 4.7.2
client/X11/xf_keyboard.c: In function ‘xf_keyboard_action_script_init’:
client/X11/xf_keyboard.c:48:6: warning: unused variable ‘exitCode’
[-Wunused-variable]
gcc 4.9
client/X11/xf_client.c: In function ‘xf_client_thread’:
client/X11/xf_client.c:1537:3: warning: ‘inputThread’ may be used
uninitialized in this function [-Wmaybe-uninitialized]
When using software gdi (/gdi:sw) the rdp update PDU callback
gdi_palette_update() is used which writes the new colors to
the rdpContext's gdi->palette buffer.
The X11 functions however access xfc->palette which gets only
updated by xf_gdi_palette_update() which is the callback if
/gdi:hw is used.
This commit changes xfc->palette to a pointer which points to
xfc's private buffer with /gdi:hw or to the gdi->palette if
software gdi is used.
1)
Added missing checks for CreateEvent which also required the
following related changes:
- changed freerdp_context_new API to BOOL
- changed freerdp_peer_context_new API to BOOL
- changed pRdpClientNew callback to BOOL
- changed pContextNew callback to BOOL
- changed psPeerAccepted callback to BOOL
- changed psPeerContextNew callback to BOOL
2)
Fixed lots of missing alloc and error checks in the
changed code's neighbourhood.
3)
Check freerdp_client_codecs_prepare result to avoid segfaults
caused by using non-initialized codecs.
4)
Fixed deadlocks in x11 caused by missing xf_unlock_x11() calls
in some error handlers
5)
Some fixes in thread pool:
- DEFAULT_POOL assignment did not match TP_POOL definition
- don't free the pool pointer if it points to the static DEFAULT_POOL
- added error handling and cleanup in InitializeThreadpool
- removed some unneeded null checks for free()
- fixed a memory leak in shadow_client
- removed rfx_compose_message_header from API
Changed the following functions to BOOL, check the result
where they are called and handle failures:
- rfx_compose_message
- rfx_compose_message_header
- rfx_write_tile
- rfx_write_message_tileset
- rfx_write_message_frame_begin
- rfx_write_message_region
- rfx_write_message_frame_end
- rfx_write_message
rfx_process_message:
- check memory allocation failures
- verify protocol-conform order of data messages to prevents memory
leaks caused by repeated allocations
- verify that header messages were parsed/received before the
data messages
- treat unknown rlgr mode as error
- fixed/added error handling
- fixed all callers to check/handle result
rfx_encode_message:
- fixed incorrect usage of realloc
- missing malloc check
- missing check of CreateThreadpoolWork
- correct cleanup on failure (threadpool, memory)
- check rfx_encode_message result
rfx_encode_messages:
- check rfx_split_message result
- correct cleanup on failure
- prevent memory leak on failure
rfx_write_message_context:
- fixed invalid channelId value (must be 0xFF for WBT_CONTEXT)
rfx_process_message_codec_versions:
- fixed invalid read size of codec_version (it is 16bit)
rfx_process_message_channels:
- verify protocol conform channelId value
rfx_process_message_region:
- replaced invalid reallocs with malloc
- read and verify regionType and numTileSets from stream
rfx_process_message_tileset:
- check allocation results
- fixed incorrect usages of realloc
setupWorkers:
- fixed incorrect usages of realloc
rfx_split_message:
- removed dead code
- missing malloc check
rfx_compose_message:
- fixed a memory leak
- check/handle rfx_encode_message result
* top level GDI functions return 0 on error and != 0 otherwise but the
low level functions (16bpp.c, 8bpp.c 32bpp.c) which are called did it
exactly the other way around. Those were adapted.
* change gdi_InvalidateRegion to BOOL and check calls where appropriate
* integrate comments from pull request
The window width/height variables and in turn SmartSizingWidth/Height
get automatically updated in th the configure notify event handler.
If the window is created initially the values don't differ from the
event values and therefore SmartSizing was not applied until the
window was resized.
- make smart-sizing work again which was killed in previous commits
- removed several unnecessary/ugly workarounds
- miscellaneous small fixes
- new feature: restore previous window position when toggling out of
fullscreen mode
- new feature: if /f is specified in combination with /smart-sizing:WxH
we run the session in the /smart-sizing dimensions scaled to full screen
Since several contributers in the past repeatedly made the error
to treat xfc->width and xfc->height as equal with the x11 window
width and heigth I've renamed these variables to sessionWidth and
sessionHeight.
Small cleanup of passing around decorations flag.
Limit PercentScreen to single monitor vs. entire desktop. IMO - this is better behavior in a multimonitor environment.
Handle fullscreen windows better:
1. Ensure that size hints are set to allow resizing before setting a window to fullscreen as some window managers do not behave properly.
2. Handle fullscreen toggles without destroying and recreating window.
3. Use NET_WM_STATE_FULLSCREEN Extended Window Manager Hint for fullscreen functionality
4. Use the NET_WM_FULLSCREEN_MONITORS Extended Window Manager Hint when appropriate
5. When a single monitor fullscreen is requested - use the current monitor(as determined from mouse location)
6. Handle cases where there is no local monitor at coordinate 0,0. The Windows server expect there to be a monitor at this location, so we maintain offset if necessary between our local primary monitor and the server side primary monitor located at 0,0.
Remove use of the visibleOffset, this completely breaks the display of all windows except for the main application window. Instead,
just maintain a local offset correction of the windowOffset.
Apply workaround to determining the workArea for remote app mode.
Since XWarpPointer generates an pointer motion event, as if the mouse
was moved by hand, xfreerdp sends back the received mouse position to
the server. This behavior is wrong since pointer positions set via
server pointer update pdu shouldn't get sent back to the server
(it's not a "real" mouse move).
To fix this problem change the x windows event mask to not handle
pointer motion events before setting the new pointer position and change
it back again afterwards.
Possible downside of this is that some pointer motion events might get
lost but this shouldn't be noticeable.
Thanks to nfedera for the neat idea ;).
The input->FocusInEvent callback implementations (normal and fast-path) have
always sent the mouse position even if the pointer was outside of the freerdp
client area. In addition xfreerdp used the wrong pointer coordinates which
were relative to the root window instead of its own.
On focus-in the pointer position must only be sent if the pointer is
currently within the program's client area. However, the clients had no way
to pass that information to input->FocusInEvent which required an API change.
- removed mouse pointer x, y parameters from input interface's FocusInEvent
- clients are responsible to call input->MouseEvent on focus-in if necessary
- fixed xfreerdp and wfreerdp accordingly
Note: /gdi:sw was working fine, this commit fixes /gdi:hw
* calculate color channel shifts based on X11 visual color masks
* fast path to skip conversion if visual color masks equal rdp color masks
* successfully tested 8/15/16/24/32 bpp rdp sessions on 16/24/32 bpp visuals
xf_cliprdr_process_selection_notify calls xf_cliprdr_send_client_format_list
if the SelectionNotify event property was None.
xf_cliprdr_send_client_format_list called XConvertSelection even if there
was no clipboard owner. In that case the XServer generates a SelectionNotify
event to the requestor (us) with property None and so on ...
The most obvious fix is to ensure that XConvertSelection is not called if
the owner is None which is done in this commit.
In case of server side initiated desktop resize it could happen that the
client was "scaling" even if smart-sizing was disabled. The reason for
this was that the "scaled" width and height was set when the X Configure
event arrived but not in xf_desktop_resize.
* fixed crash on server-initiated desktop resize and fullscreen toggle with /gdi:sw
* implemented desktop resize for /gdi:sw
* check for availibility of XRender extension and show error messages if a setting would require it
* fixed compile error if xrender devel packages are missing
* fixed bug with server-initiated desktop resize in fullscreen mode
Most of xfreerdp's hw gdi callbacks did the the same X11 drawing operation
twice: Initially on the primary pixmap to keep the back buffer up to date
and then directly on the window in order to see the result.
This sequence is followed by a call to gdi_InvalidateRegion() to mark the
rectangular area of the drawing operation.
Since in xfreerdp's current design (or better current evolved state) the
backbuffer pixmap is obligatory, all drawing operations directly targeting
the window are redundant because of the subsequent EndPaint (mapped to
xf_hw_end_paint) where the damaged region is copied from the backbuffer
to the drawable.
- removed X11 drawing operations which directly target the window drawable
- fixed some errors in the calculation of the required update regions
X11 gdi polyline has been broken by commit 3de5c62.
We don't need to convert the point coordinates if XDrawLines is called with
coordinate mode CoordModePrevious.
The fOpRedundant field of the GlyphIndex primary drawing order
(MS-RDPEGDI, chapter 2.2.2.2.1.1.2.13) was neglected which resulted in some
severe text rendering errors.
"libfreerdp" consisted of multiple (small) single libraries. If the cmake
option MONOLITHIC was used only one library was build combining all of
the libfreerdp-* libraries.
The only exceptions to this are libfreerdp-server and libfreerdp-client these
are build as separate libraries.
This commit obsoltes non-monolithic builds and makes monolithic builds
the default. The cmake option MONOLITHIC is also removed.
Conflicts:
channels/drdynvc/client/dvcman.c
include/freerdp/codec/h264.h
libfreerdp/codec/h264.c after this merge h264 doesn't work anymore!!
libfreerdp/utils/svc_plugin.c
* client/X11/xf_client.c (xf_pre_connect): Set the field `colormap' of
`struct xfContext' to the default colormap.
* client/X11/xfreerdp.h (xf_gdi_get_color): Declare new external function.
* client/X11/xf_gdi.c (xf_gdi_get_color): Define new external function.
Use xf_gdi_get_color() to get the appropriate X color index from the GDI
color representation.
* client/X11/xf_graphics.c: Likewise.
representation.
* client/X11/xf_gdi.c:
Use freerdp_color_convert_drawing_order_color_to_gdi_color() to convert
from drawing order color representation to GDI color representation
troughout.
* client/X11/xf_graphics.c: Likewise.
* Zeroing xevent helped address some erratic behavior.
* valgrind complained about using xfBitmap uninitialized
during shutdown, traced it back to the initialization.
Bitmap_Prototype->size > sizeof(rdpBitmap).
* Early exit from recv_tpkt_pdu is necessary to address
a shutdown crash - the channelId value was being used
without being set in the disconnect case.
winpr is now always build as single library.
The build option MONOLITHIC_BUILD doesn't influence this behavior anymore.
The only exception is winpr-makecert-tool which is still build as extra
library.
This obsoletes complex_libraries for winpr.
The X11 core protocol does not have support for selection ownership
notifications. Until now xfreerdp worked around this issue by always sending
a format list pdu to the server after sending the format data response pdu
which makes the server side think that the clients clipboard data has changed.
This workaround has some severe drawbacks:
* it causes unnecessary data transfers because even without local clipboard
data changes the same data is always re-transferred over the channel
* with some clipboard managers (in the server sessions) you will get massive
endless data transfer loops because these managers immediately request the
data on clipboard changes.
The correct (core X11) way would be polling for selection ownership changes
which must include the ability to detect changes to the TIMESTAMP target if
the selection owner did not change.
The alternative to the poll based approach is using the X Fixes extension in
order to get selection ownership notifications.
This commit adds support for the XFIXES solution and also moves the complete
clipboard related event handling from xf_event.c to xf_cliprdr.c
ilammy