of locking before processing the message (single/all window lock)
-> in most message cases, I could comment out the unlocking/locking which
switched to the different lock, because the required lock is now already held,
this removes some race conditions which were commented in the code already
* EventDispatcher::SetDragMessage() didn't lock the object, this would have
been bad if multiple windows tried to set a drag bitmap at once
* the Desktop object keeps track of mouse position and pressed buttons, so
that it doesn't need to lock the EventDispatcher for sending fake mouse
moves to windows on show/hide of windows (solves some cases of possible
dead locks with the new locking strategy)
* the keyboard EventFilter switches the current workspace asynchrnously from
the Desktop thread (another source of possible deadlocks)
* the "reader is trying to become writer" check in MultiLocker is only used
in DEBUG mode now
As a summary: It would be nice if ServerWindow used a readlock for all messages
it processes itself, and forwards all messages for which it needs a write lock
to the Desktop thread. All cases where either the Desktop or the ServerWindow
mess with the EventDispatcher are possible sources of deadlocks. This is solved
right now by making sure that the lock is released before using the
EventDispatcher.
I have not observed any deadlocks while switching workspaces and launching
many apps anymore, neither crashes. But I have not tested extensively except
for in the test environment. That being said, I could reproduce the problems
on first try before in Haiku.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22410 a95241bf-73f2-0310-859d-f6bbb57e9c96
possibly making them a little faster too
* mess with decorator button size calculation to make the whole layout scale
more agreeable with the font size (no more fixed offsets/insets), but it
is work in progress
* DefaultDecorator no longer allocated the border color array, it is part of
the object now
* small memory footprint optimizations in ViewLayer, Decorator and WindowLayer
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22003 a95241bf-73f2-0310-859d-f6bbb57e9c96
of the active ViewLayer is now always mirrored in the Painter instance
of a ServerWindow, so that it doesn't need to be synced on every drawing
command, this was previously incomplete for font handling
* removed the DrawState parameter from all the DrawingEngine functions
* adjusted ServerWindow and ServerPicture accordingly
* made sure that string related functions used by non-drawing related
parts (ServerApp, Decorator) don't interfere with the current drawing
state
* moved AS_SYNC handling from _DispatchViewMessage to _DispatchMessage,
it is actually a window message and doesn't require fCurrentLayer to
be valid
* fixed bug #1300, fCurrentLayer was not updated when a ViewLayer was
deleted by client request which happened to be fCurrentLayer (I am now
handling it so that the parent becomes the current layer, could be
wrong)
* AGGTextRenderer is no longer using it's own scanline, which should save
a few bytes RAM, the Painter already had such an object
* StringWidth() in AGGTextRenderer is now taking the escapement_delta into
account
* Painter::StrokeLine() doesn't need to check the clipping as much, since
that is already done in DrawingEngine
* if a ServerWindow message is not handled because fCurrentLayer is NULL,
a reply is sent in case the messages needs it (client window could
freeze otherwise, waiting for the reply for ever)
* removed unused AS_SET_FONT and AS_SET_FONT_SIZE
* added automatic RGBColor -> rgb_color conversion to RGBColor.h
* minor cleanup for 80 char/line limit
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21685 a95241bf-73f2-0310-859d-f6bbb57e9c96
I "ported" the region implementation from XOrg to work on BRegion data.
This resulted in pretty much the same code structure as before, with
RegionSupport.cpp containing the messy details. Only now it _is_ really messy
from a code beauty point of view. I didn't exactly feel like cleaning it
up right now... but I guess I will have to.
So what does this mean - our BRegion implementation was very slow (no offense!),
and on top of that it scaled very badly with more and more rects. The new
implementation seems to be on par with the very fast R5 implementation and
the data looks exactly the same too. BRegion is very performance critical
for the app_server, and I cannot wait to try this on my slow computer...
Some changes are noteworthy: The right and bottom coordinates of
BRegion internal data are now exclusive! I inherited that from the
XOrg implementation and didn't feel like changing the code, seeing it
is probably tested quite well. The conversion is handled transparently.
Secondly, constructing a BRegion with just one rect is not invoking
malloc anymore for the member data, this makes it much more efficient
to use temporary BRegions with just one rect, both externally and internally
in the BRegion implementation.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21665 a95241bf-73f2-0310-859d-f6bbb57e9c96
a string
* fixed profiling of message processsing in ServerWindow (didn't take batch
processing into account)
* accelerated ViewLayer::RebuildClipping() by a factor of two by avoiding
BRegion::Exclude(clipping_rect) for each child, and instead building
one region with all children, and excluding that. RebuildClipping() is
quite a common operation and is quite slow for views with many children
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21646 a95241bf-73f2-0310-859d-f6bbb57e9c96
of BMallocIO. Added an additional constructor to handle a file. This is
in preparation of implementing BView::SetDiskMode().
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@20996 a95241bf-73f2-0310-859d-f6bbb57e9c96
* test the actually visible view area against dirty region,
(this avoids making the client window call the Draw() function of views,
even though all the drawing would get clipped on the server side)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@20756 a95241bf-73f2-0310-859d-f6bbb57e9c96
Removed most data allocations/copying from PicturePlayer, ServerPicture now has to do this when converting coordinates.
Added additional functions to ViewLayer to copy&convert multiple BPoint, BRect, BRegion to Screen coordinates, those should be further optimized.
Removed some function call overhead.
Note: some functions of PicturePlayer don't appear to be implented by PictureDataWriter,
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@20292 a95241bf-73f2-0310-859d-f6bbb57e9c96
The problem was that the view's screen clipping was not updated if its frame did not change
because of a resized parent - but that might be needed if the new parent frame reveals a new
portion of that view.
I added a TODO so that if there is a way to test for this case, we only need to invalidate
the clipping if really needed. For now, we always do it.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@19695 a95241bf-73f2-0310-859d-f6bbb57e9c96
LockExclusiveAccess() (meaning more or less access to the
frame buffer)
* extracted the AGGTextRenderer to be a global instance used
by each Painter instance (currently, it is thread safe because
of the global font lock, so there is some work left in this
regard)
* gave every ServerWindow it's own DrawingEngine instance, this
is work in progress. So far, there doesn't seem to be a regression,
but less fighting over the exclusive access to the frame buffer, now
each ServerWindow thread can draw in parallel. There is room for
improvement, plus I think I'm leaking the DrawingEngine...
* changed the locking for the software cursor. ShowSoftwareCursor()
can only be called if HideSoftwareCursor(BRect) returned true, or
if you called the generic HideSoftwareCursor(), since it needs
to keep the cursor lock and unlocks in Show...!
* some clean up and renaming in Decorator and friends
* moved PatternHandler.h to live along with the .cpp
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@19427 a95241bf-73f2-0310-859d-f6bbb57e9c96
changes in the HWInterface (ie on mode switch)
* initialization and shutdown of the HWInterface instance no longer
go through DrawingEngine, which had nothing to do with it in the
first place
(this is in preparation of giving each ServerWindow it's own
DrawingEngine instance)
* small performance improvement in ViewLayer::ScrollBy()
* some cleanup in ServerConfig.h
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@19391 a95241bf-73f2-0310-859d-f6bbb57e9c96
but use integer coordinates. These are now used in ViewLayer for the
coordinate system (layout, scrolling offset, view bitmap layout)
* modest performance improvements by inlining some very often used
methods, and by preventing to go down the entire layer tree too often,
for example InvalidateScreenClipping was always called in the deep
mode, therefor it is save to assume that the tree does not have to
be traversed as soon as the clipping is already invalid
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@19390 a95241bf-73f2-0310-859d-f6bbb57e9c96
* fixed SetViewColor() - I have no idea why it compiled before
* changed SetHidden() to make less of a difference between
the top view and other views, the clipping is now always
rebuilt
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17527 a95241bf-73f2-0310-859d-f6bbb57e9c96
it is added our removed to another view, this is
taken care of in View.cpp until we find something
better
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17511 a95241bf-73f2-0310-859d-f6bbb57e9c96
ViewLayer::SetViewBitmap() did not show the overlay, only updated it.
* Simplified overlay handling a bit, removed Overlay::Show(), and IsVisible(),
replaced Update() by Configure().
* Made similar changes in the HWInterface as well.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17504 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Overlay::SetView() no longer calls HWInterace::UpdateOverlay() if it's currently
hidden.
* ViewLayer::UpdateVisibleDeep() now calls _UpdateOverlayView() before showing
the overlay.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17489 a95241bf-73f2-0310-859d-f6bbb57e9c96
the overlay buffer changed. Found by Marcus.
As a result, the overlay window looks much smoother when moving it around (and it
even starts to move when you don't change the overlay bitmap at all...).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17484 a95241bf-73f2-0310-859d-f6bbb57e9c96
children of the view in question into account. Neither
are they included in the region being copied, nor are
views overdrawn by the copied contents. Also, the update
region is calculated from the region having been copied,
so that holes from children views are invalidated
correctly.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17268 a95241bf-73f2-0310-859d-f6bbb57e9c96
source file.
* An overlay is now also hidden in case its is removed from the window.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17209 a95241bf-73f2-0310-859d-f6bbb57e9c96
is currently broken, mode switches probably fail or result in sudden death
(didn't try) it's good enough for Radeon cards and VLC (might work with
others as well).
* Implemented follow modes for view bitmaps (wasn't taken into account at
all before).
* Switched to a darker overlay color for now (dunno what exactly makes a
good candidate there, but this one is good enough for now).
* Added TODO about a race condition in AS_LAYER_SET_VIEW_BITMAP.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17208 a95241bf-73f2-0310-859d-f6bbb57e9c96
is managed for those bitmaps:
- the shared client memory mechanism is used to allocate a small overlay_client_data
structure that contains the actual buffer and a semaphore that you have acquire in
order to access it.
- LockBits()/UnlockBits() now have a function: you need to call them before accessing
the overlay buffer, and you need to keep that lock until you're done with it.
* The overlay cookie is now an extra member of the ServerBitmap class.
* Removed fInitialized from ServerBitmap - IsValid() now just checks the buffer associated
with the bitmap.
* ViewLayer::Draw() will now handle overlay bitmaps specially and will draw the overlay
color instead of any contents (this is currently in ugly pink, but will become some
dark color later on).
* All what's missing from actually being able to use overlays now is to configure
them so that they are shown on screen. VLC will now show an empty pink window when
overlay video is enabled.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17201 a95241bf-73f2-0310-859d-f6bbb57e9c96
ViewLayer (for example, fixes NetPositive rendering
HTML with a background image)
* use BRegion pool everywhere in ViewLayer
* WindowLayer update sessions distinguish between
different reasons for the update: exposed and requested -
on expose updates, the view backgrounds are cleared
immidiately (as on R5), to keep the time previous stuff
keeps showing as short as possible, while on requested
updates, the background clearing is delayed until the
client draws something, to keep the time until the client
fills a view with content as small as possible to reduce
flickering (might need more work, could be buggy yet)
* HWInterface and DrawingEngine support delayed syncing to
the graphics hardware at least for FillRect/Region. The
speed up gained by this is minor though.
* HWInterface cursor rendering uses a bit of rounding to
avoid the slight transparent shadow around the cursor
(I don't know if it is fully correct though, at least the
shasow disappeared)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17172 a95241bf-73f2-0310-859d-f6bbb57e9c96
to cut down on BRegion related allocations, cannot really tell
if it speeds things up
* used the new BRegion pool in WindowLayer and ViewLayer whereever
a BRegion was used on the stack
* fixed the debugging stuff in MultiLocker - it will get you into
the debugger if you
- try to nest read locks
- try to write lock when your are a reader already
- don't match up nested locks when your a writer
-> but only if you #define DEBUG 1 in the .cpp, is off by default now
* went over WindowLayer, ServerWindow, Desktop and a few other places
and fixed the locking for use with the MultiLocker, the "a reader can
not become a writer" is especially tricky, feel free to review the
changes
* activated the MultiLocker, I tested this quite a bit, if there are
problems simply turn on DEBUG and you should drop into the debugger
right where the problem is... hope all is good
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@17046 a95241bf-73f2-0310-859d-f6bbb57e9c96
This lightens the problem in bug #98 a bit, but doesn't completely fix it; you
still don't see any items in that list on the switch, but they now appear again
when you scroll around there.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@16997 a95241bf-73f2-0310-859d-f6bbb57e9c96
can't move them to another workspace, and you currently don't see a
window moving that is not in the current workspace. (this fixes bug #135)
* Desktop::SetWindowBehind() didn't update the WorkspacesLayer.
* Changed Desktop::MoveWindowBy() to be able to move window on any workspace.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@16776 a95241bf-73f2-0310-859d-f6bbb57e9c96
take an empty region as an indication to remove the clipping, but on R5,
this is actually valid... this patch fixes the problem
* the ViewState::clipping_region is now consistently used to memorize
the user defined clipping of the view state instead of being sometimes
used to cache the current view clipping
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@16656 a95241bf-73f2-0310-859d-f6bbb57e9c96
* all cursors owned by a team are visually different,
or (iaw) an already existing cursor is reused when
it is set by the client again
* changed various occurances of cursor data from "int8*"
to "uint8*"
* ServerCursors also remember the R5 data from which
they were created
* the reference counting and destruction of
ServerCursors changed: The cursor knows it is attached
to a CursorManager and one can simply use
ServerCursor::Acquire() and Release() and the reference
counting and everything is being taken care of
* destroying a ViewLayer will now correctly release a set
ServerCursor
* fixed a race condition when setting a cursor through
BView::SetViewCursor(): If the client code looks like this:
BCursor cursor(cursorData);
someView->SetViewCursor(&cursor, false);
there is a relatively high chance the BCursor destructor
told the ServerApp thread to destroy the cursor before
the ServerWindow thread got to "acquire" the cursor for
use by the view layer. The very same problem is likely the
reason that SetViewCursor works to unreliably on R5, even
when the "sync" flag is set to "true" (although it should
theoretically work in that case).
all these fixes make WonderBrush work fine again with the
new support of custom cursors.... coded by axeld and myself
(the joys of pair programming :-)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@16521 a95241bf-73f2-0310-859d-f6bbb57e9c96
omitted. This fixes bug #181, or at least the part of it I can reproduce.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@16442 a95241bf-73f2-0310-859d-f6bbb57e9c96
Should probably be cleaned up a little. This fixes bug #146.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@16377 a95241bf-73f2-0310-859d-f6bbb57e9c96
* speed up from/to screen conversion of BRects and BRegions
* invisible views don't need updating
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@16257 a95241bf-73f2-0310-859d-f6bbb57e9c96
work though, as HWInterface can only draw B_RGB32 cursors...
* More build fixes for libbe_test target - it defines __HAIKU__ as well, now
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@16211 a95241bf-73f2-0310-859d-f6bbb57e9c96
* fixed weird pointer conversion in SetStyle()
* fixed a potential mix up in operator=() in case the
other ServerFont has fStyle == NULL
ServerWindow:
* the WindowLayer fTopLayer cannot be deleted by
client request, just for safety reasons
* the link is flushed if there is no drawing engine,
but this case is theoretical only
* deleting the ServerWindow object syncs with the
client, so that when BBitmaps are deleted, they
can be sure there are no pending messages (which
would be executed in a nother thread)
* there is no timeout anymore when sending messages
to the client, which made absolutely no sense
AGGTextRenderer:
* renamed fFontManager to fFontCache, because that's
what it really is
* fLastFamilyAndStyle defaulted to the system plain
font and therefor that font was never loaded when
the font never changed meanwhile
DrawingMode:
* I'm not quite sure but I think there was the
potential of a division by zero, at least I
had crashes with "divide error"
HWInterface:
* fix update when the cursor shape changed in
double buffered mode
ViewLayer:
* since the top layer is never really deleted
before its time has come, it is not necessary
to set it to NULL in the ViewLayer destructor
ViewLayer/WindowLayer:
* added a function to collect the view tokens
that are affected by an update session
EventDispatcher:
* use the importance of the message for the timeout
in _SendMessage()
* drop mouse moved events in the server if we're
lagging behind more than 5 ms (Axel, maybe review)
View:
* there were some problems with the locking
of the BWindow looper in RemoveSelf(), since
this is called from the window destructor,
also of BWindows from BBitmaps, which have
never been run (this might need review), at
least I seem to have solved the crashing
problems introduced by actually deleting the
view hirarchy in the BWindow destructor
* fixed _Draw() for being used non-recursively,
temporarily disabled DrawAfterChildren, which
didn't work yet anyways (because views cannot
draw over children in the server yet)
Window:
* small cleanup when deleting shortcuts
* sync with the server when having send
AS_DELETE_WINDOW (see ServerWindow above)
* fixed locking in Begin/EndViewTransaction()
* removed folding of _UPDATE_ messages, since
there is only one ever in the queue
* set the fInTransaction flag during an update,
I plan to use this in BView later to
flush the link when drawing outside of an
update
* BView::_Draw() is now called by view token,
this gives the next leap forward in speed,
the overhead because of drawing clean views
was considerable
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15878 a95241bf-73f2-0310-859d-f6bbb57e9c96
* make sure layers are removed from the token space when the
WindowLayer is deleted
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15852 a95241bf-73f2-0310-859d-f6bbb57e9c96
it could crashed the server.
* ViewLayer now deletes the view bitmap on destruction, if any.
* BitmapManager::Delete() now also accepts NULL bitmaps.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15723 a95241bf-73f2-0310-859d-f6bbb57e9c96
or even the resizing mode isn't done yet, though. See TODOs.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15720 a95241bf-73f2-0310-859d-f6bbb57e9c96
