There was some mixup with the interrupt registers, still:
- The driver uses 16-bit read/write, but on SandyBridge the register is
32 bits
- There is a global interrupt enable bit, which must be set to unmask
everything else
- The bits for vblank interrupt are not the same on SNB and later PCH
based devices, and the code mixed the two.
Move the computation of the interrupt bits to an helper function, and
use it everywhere to make sure we always use the right bits.
Modesetting
===========
My previous hack was setting the transcoder registers, instead of the
display ones. Do that the way it is designed in the driver instead:
- If there is a transcoder, set its registers, but do not set the
display timings. The display will remain set at its native (and only)
resolution, and panel fitting will adjust the output of the transcoder
to match.
- If there is no transcoder, set the display registers directly to the
native resolution, as it was done on previous generation devices.
- fPipeOffset hacks no longer needed
DPMS
====
It seems the panel control register is not readable on PCH? Anyway, the
code would loop forever waiting for the bit to become unset when turning
the display off. Waiting seems to not be needed, so just remove it as
well as the "unlock" bit, which does not work for me and results in a
black screen.
Remaining hacks
===============
I still need to force HEAD_MODE_A_ANALOG to get output on pipe B (LVDS
display) working. I suspect something is common to the two pipes or not
allocated to the right one.
This version will have less side effects on other generations and help
with getting things to work on SandyBridge and possibly later devices.
Please test and report.
BeOS had a feature where you could click on a list and while holding
down the mouse button scroll through the list items selecting them
as you went. I implemented the same feature on Haiku.
Did the work in separate thread which makes sure that the mouse
is held down the whole time even if you leave the view.
Thanks Diver for bringing this feature to my attention.
* Rename kDoubleClickTresh to kDoubleClickThreshold
and use floating point in rval
* Pointer (*) goes with type for property var
* Use {}'s for multi-line if conditional
* rename point to where (consistent with header and rest of IK)
* Explicitly compare with 0 for bitwise operator statements
* Rename InitiateDrag() params
- This is how it is named in other versions of elf.h (Linux, glibc, possibly more)
- ELF_MAGIC is used by libelf for the same thing, and the defines conflicts,
breaking libelf build on Haiku.
Add an Icon() and SetIcon() method. Override SetMarked() to set the
parent menu field icon.
Don't move the label right if icon is NULL. Make the first menu item
an IconMenuItem with a NULL icon. This allows the icon to draw in the
closed menu state.
icon gets updated even if you select an item in a submenu
http://insightfactory.tumblr.com/image/142366356207
* Make the color box a rectangle with proportions of golden ratio.
* Override GetContentSize() to make menu item area larger.
* Label should never truncate since I make sure there is enough room.
* Draw the label using BMenuItem parent class
* Carefully adjust the spacing so that there is an attractive amount of
padding between the checkmark and color box and the color box and label.
Add _AddMenu method to BMenuField that adds BColorMenuItem as its
base menu item. This shows the BColorMenuItem in the closed state.
Create BPrivate::MenuItemPrivate
Add a SetSubmenu() method to MenuItemPrivate that gives you
the ability to add a submenu after creating the object. This
method should be public
Skip disabled items
Color gets updated even if you select an item in a submenu
Architecture:
- Store and provide accessor for the size in bytes of the low-level
debug_cpu_state size of the respective target CPU. Adjust subclasses
to pass in the appropriate size information.
The only file in here was Deskbar.h, which just included the "real"
Deskbar.h. Considering nothing in-tree cares about this, and only
a small number of applications at HaikuArchives do, and that Haiku already
broke BeOS source compatibility in lots of other ways, let's just remove
this deprecated cruft.
The warnings made those headers and those including them not detectable
by libiberty's configure because of the way it checked for them.
This fixes the gcc build.
* Detect PCH model based on ISA bridge and save
into shared info for later use.
* On CougarPoint PCH systems, assign pipes via
special CPT registers
* Drop HasPlatformControlHub as PCH should be
based on more than just generation.
UserInterface:
- Add Clone() function to set of required virtuals. This asks the subclass
to create a new instance of its respective type.
{CommandLine,Graphical,Report}UserInterface:
- Implement the above function.
TeamDebugger:
- Add accessor for the currently active UI.
TargetHostInterface:
- Set correct request type when setting up the options for a team restart.
- Ask the TeamDebugger for its user interface and clone it in order to fill
in that aspect of the debug options. This fixes a regression introduced in
commit 880a64, which inadvertently resulted in team restarts no longer
working.
* Wait evaluation is done considering both position
and buffer size.
* Set back position to 0 after seeking.
* Simplify EvaluatePosition to take advantage of inheritance.
* Use back position to determine the current size of the buffer.
* Add more debug tracing.
* Other minor fixes.
* Implement Open/Close mechanism.
* Implement SetBuffer.
* Implement timeout handling.
* Improve Seek mechanism, this is now working by locking the
thread until the backend call SeekCompleted.
* Moved data locking in the internal buffer class, includes handling
of backend writes.
* Takes advantage of inheritance for getting the size. The choice is
made depending on the flags. This allow the implementer to easily
return a custom size by reimplementing GetSize. At the same time
a plain BAdapterIO can still have it's total size set, but the behavior
will change depending it's mutable or not.
* Some decisions are now made by considering everything in absolute
values.
* Other minor fixes.
Fixes#12710.
Signed-off-by: Augustin Cavalier <[email protected]>
I fixed the modifications to the Jamfiles in src/bin, they were all wrong
in the patch.
Signed-off-by: Augustin Cavalier <[email protected]>
Fixes#12575. I checked with PulkoMandy on IRC before merging this,
and fixed some whitespace violations of the coding style.
* This is by default provided using a relative buffer. When the
client request a seek over the range we currently have, we will
ask the backend with a SeekRequested. When the backend operation
ended successfully, the client is required to call SeekCompleted
before to return. This will cause the offset to change and the
buffer is cleaned as result. Next data received will be considered
to be at a position equal to the offset plus the relative position.
Adrien Destugues