inline const floats are a gcc extension. It is possible to do it in a
standard way in C++11 using constexpr, but then gcc will reject the
previously accepted nonstandard syntax.
* ... instead of queuing it for the job thread. The advantage is that
the request will be handled immediately and clients won't have to wait
for transactions (which may even require user feedback) to finish. It
complicates Volume a bit, since there are now two threads that may
access it. The shared data have been moved to a State object which is
protected by a lock.
* For commit transaction requests check whether another package request
is already pending/in progress before queuing a job. Fail immediately,
if there is.
Fixes bug #10039.
Pass the partition size into the default name, not the device size,
this makes the variable names match the comment.
There is no functional change here, the variable names got renamed
to make it more clear what's happening. We want show the partition
size, not the device size, and we want to error if the device size is less
than the partition size, not if the partition size is less than the device size.
and not any other modifier such as command or control or shift.
Fixes#10511
We need to be more strict here because Tracker defines Cmd+Opt+Up and
Cmd+Opt+Down shortcuts, S&T shouldn't override them.
Clipping paths (or rather alpha masks) and affine view transformations
seem to be generally supported for all drawing operations now. There might
still be bugs that have not yet been exposed by the current set of tests.
These could be related to scrolling, origin and scale as well as stacked
view states.
* Turn off optimized code-paths when an alpha mask is set. The generic
path already included the alpha-mask version of the AGG pipeline.
* Add transformation to generic bitmap rendering. (It already worked
with a transformation, so there is no slow down compared to before,
only now it also includes the transformation set via BView::SetTransform())
The initial core assignment has to be done without any knowledge about
the thread behaviour. Moreover, short lived tasks may spent most of their
time executing on that initial core. This patch attempts to improve the
qualiti of that initial decision.
The main purpose of this patch is to eliminate the delay between thread
migration and result of that migration being visible in load statistics.
Such delay, in certain circumstances, may cause some cores to become
overloaded because the scheduler migrates too many threads to them before
the effect of migration becomes apparent.
In order to keep the scheduler tickless core load is computed and updated
only during various scheduler events (i.e. thread enqueue, reschedule, etc).
The problem it creates is that if a core becomes idle its load may remain
outdated for an extended period of time thus resulting in suboptimal thread
migration decisions.
The solution to this problem is to add a timer each time an idle thread is
scheudled which, after kLoadMeasureInterval, would fire and force load
update.
I didn't mean to omit this check, I just need to save deviceSize below for
get_default_volume_name() so that it will match the mount point name.
Checked to make sure the calculation works on a 2GiB, 32GiB, and 128GiB
(virtual) exfat volume.
* Set the size of the volume name member var to B_FILE_NAME_LENGTH which is the
max width of a volume name according to the BeBook and the Support Kit code. We'll
deal with trying to stuff the volume name back into the exfat volume label later.
* Rename volume_name() to get_volume_name()
* Pass the name string length into get_volume_name() and use it to avoid buffer overrun.
* Fill name with a blank string if volume has no label.
* Don't memset the name with zeros before doing the conversion from Unicode to UTF-8,
the conversion function will make sure the result is NUL-terminated if it returns B_OK.
* Also check the return value of the Unicode conversion function and return an error if it fails.
* Add get_default_volume_name() method to Utility.cpp which is used to fill out the default
volume name in the case volume name is blank. e.g. 32GiB ExFAT Volume. This now
applies to both the volume name and mount point.
* Use non-metric prefixes for default volume name, e.g. MiB, GiB, TiB
* For an unset volume name fill the volume name with an empty string.
* Remove the leading underscore from _name and _partition parameters
* Replace size constants with sizeof() calls
* Remove Axel from the copyright statement in Utility.cpp, he had nothing to do with it,
add Jérôme (aka korli) because he wrote the code for get_default_volume_name()
(was in Volumes.cpp)
* Remove some trailing spaces from encodings.cpp
Thanks Axel.
* build kernel libgcc and libsupc++ with disabled posix threads for all
architectures. We currently change the configuration manually, as gcc
doesn't easily let us reconfigure without a full rebuild.
Fixes#10501
* update copyright headers, attribute to Haiku, Inc. add authors
* volume_name gets filled out by a utility function in Utility.cpp
* update exfat_entry
- rename name_label to just label
- adjust volume_label to have 11 uint16 chars plus 8 uint8 reserved
- add guid partition info
* Added a couple new entry type defines
* fName is 34 bytes long which fits the 11 3-byte UTF-8 chars and a \0
* The maximum number of allowed ranges is now taken into account.
* Still untested on my end, however, I tested the array copying code with
a number of test data.
* Also removed wonky BeOS rendering of stroked round rects and ellipses.
Nobody would expect it to work like this. This affects stroked round rects
and ellipsis with a pensize greater than 1.
* Refactored common code from _StrokePath() and _FillPath().
* _StrokePath() returned a curious bounding box that didn't take into
acount the miter width. Now the bounding box is computed from the actual
stroke conversion of the path.
Everything untested, but compiles, so it must work. The idea is to introduce
BAffineTransform additionally to the existing Origin and Scale properties of
BViews. One may use it in parallel or as an alternative. Painter in app_server
is not yet aware of the additional transformation. It is however already used
to transform drawing coordinates. It probably needs to work differently,
perhaps only in Painter and AGGTextRenderer.
* This is less pretty, but we need access to the connector
to find the HPD gpio pin mask on the card.
* dp_aux communications seem to work again.
* If you have a DisplayPort item attached to your card you
may want to just unplug it at this point. We attempt DP
link training and it fails. This failure will also cause
other monitors to not function as app_server still isn't
multi-head aware (#10486)
There wasn't an out-of-bounds access, I read the code wrongly. Keep the code
in sync with the AGG version. What happens is that for x < 0 and xmax < 0,
the code will memset the x < 0 region twice, unless I am still irritated...
* The alpha mask is no longer stored with 75% more memory than needed,
resolving a TODO.
* AlphaMasks are now BReferenceables.
* AlphaMasks are transferred to a pushed DrawState.
* When an AlphaMask is set on a DrawState, it sets the previous state's
mask to the AlphaMask. That one now takes care of updating not only itself,
but the previous mask as well (which works recursively).
* In AlphaMask::Generate(), a combination happens with the previous state's
mask, which again works recursively in case the previous mask also needs
to be updated. This step is combined with extracting the alpha channel from
the UtilityBitmap used to play the ServerPicture.
* Fixed some out of bounds access to memory in the "outside" case in
agg_clipped_alpha_mask.h. This happened when the requested region was
both before and after where the mask has data.
* The transaction_type(.., ..., SOLVER_TRANSACTION_RPM_ONLY)
categorization of the transaction steps is incorrect since rpm
automatically replaces an older package version when a new one is
installed and thus such package removals would be omitted. Use a
simpler check instead.
* Omit the unnecessary explicit check whether a package really has to be
installed.
Thanks to Michael Schroeder @ SUSE for his help figuring this out.
This is required on OS X and other systems which do not have the glibc
extensions for regular expressions (FreeBSD is not one of them as it
already includes gnuregex in /usr). With this there are no hardcoded
non-standard paths for OS X anymore.
regex.c and regex.h are from the official gnuregex 0.12 distribution,
the only modification is that I added __BEGIN_DECLS and __END_DECLS to
regex.h.
* Bump i2c data into it's own struct
* Populate gGPIOInfo with general gpio pins
(for things like HPD, and voltage control)
* Attach HPD gpio pins to connectors
* Dump trace info of HPD pins as well as i2c pins
* There are GPIO pins for hotplug interrupts and i2c
communications (for edid). Add two fields to connector
(index for each in gGPIOInfo).
* DP aux communcation seems to want the hotplug GPIO pin
info, which we don't get (I think this is why DP AUX comms
is broken at the moment)
* GPIO LUT seems to know about a wide range of GPIO pins, while
the i2c gpio calls only know about the ones needed for i2c.
* I'm tempted to populate gGPIOInfo with the LUT pins and
then suplement that data with the i2c GPIO pins data.
(however, not sure how many pins there are generally defined
so it could impact performance. More investigation is needed)
This would reverse how it works now with this commit.
Priority penalties were made more strict in order to prevent situation
when two or more high priority threads uses up all available CPU time
in such manner that they do not receive a penalty but starve low priority
threads.
However, a significant change to thread priorites has been made since and
now priority of all non real time threads varies in a range from 1 to
static priority minus penalty. This means that the scheduler is able to
prevent thread starvation without any complex penalty policies.
Originially, core load was a sum of eastimated loads of all currently
running or ready threads on a given core. Such value is changing very
rapidly preventing the thread migration logic from making any reasonable
decisions.
This patch changes the way core load is computed to make it more stable
thus improving the qualitiy of decisions made by the thread migration logic.
Currently core load is a sum of estimated loads of all threads that have been
ready during last load measurement interval and haven't been migrated or
killed.
The main reason for this patch is to fix gcc 4.8.2 warning about
hierarchyLevels possibly being used not initialized. Such thing
actually can not happen since all x2APIC CPUs are aware of at least
3 topology levels. However, once more topology levels are introduced
we will have to deal with CPUs that do not report information about all
of them.
* The DPMS hook would be called on all
possible displays causing lots of link
training on incomplete connector structures
that may be not attached / connected.
