|author||dscho <dscho>||2001-10-10 23:56:04 +0000|
|committer||dscho <dscho>||2001-10-10 23:56:04 +0000|
copyrect corrections, fd_set in rfbNewClient, dox in rfb.h for pthreads
Diffstat (limited to 'CHANGES')
1 files changed, 18 insertions, 3 deletions
@@ -1,4 +1,14 @@
+ rfbNewClient now sets the socket in the fd_set (for the select() call)
+ when compiling the library with HAVE_PTHREADS and an application
+ which includes "rfb.h" without, the structures got mixed up.
+ So, the pthreads section is now always at the end, and also
+ you get a linker error for rfbInitServer when using two different
+ pthread setups.
+ fixed two deadlocks: when setting a cursor and when using CopyRect
+ fixed CopyRect when copying modified regions (they lost the modified
+ WIN32 target compiles and works for example :-)
fixed CopyRect (was using the wrong order of rectangles...)
should also work with pthreads, because copyrects are
always sent immediately (so that two consecutive copy rects
@@ -6,11 +16,16 @@
changed rfbUndrawCursor(rfbClientPtr) to (rfbScreenInfoPtr), because
this makes more sense!
flag backgroundLoop in rfbScreenInfo (if having pthreads)
- rfbCopyRect & rfbCopyRegion (really copies rectangle in frameBuffer)
- added flag to optionally not send XCursor updates.
+ CopyRect & CopyRegion were implemented.
+ if you use a rfbDoCopyR* function, it copies the data in the
+ framebuffer. If you prefer to do that yourself, use rfbScheduleCopyR*
+ instead; this doesn't modify the frameBuffer.
+ added flag to optionally not send XCursor updates, but only RichCursor,
+ or if that is not possible, fall back to server side cursor.
+ This is useful if your cursor has many nice colours.
fixed java viewer on server side:
SendCursorUpdate would send data even before the client pixel format
- was set.
+ was set, but the java applet doesn't like the server's format.
fixed two pthread issues:
rfbSendFramebuffer was sent by a ProcessClientMessage function
(unprotected by updateMutex).