* avoid mixing input and output paths when building a tree. In fact, an
audio input shouldn't use only a mixer used for output.
* in case an audio input has more than one input, use these inputs to
build a mux control.
* should help with #8270, #8333
The lowest 4 bits of the MSR serves as a hint to the hardware to
favor performance or energy saving. 0 means a hint preference for
highest performance while 15 corresponds to the maximum energy
savings. A value of 7 translates into a hint to balance performance
with energy savings.
The default reset value of the MSR is 0. If BIOS doesn't intialize
the MSR, the hardware will run in performance state. This patch
initialize the MSR with value of 7 for balance between performance
and energy savings
Signed-off-by: Fredrik Holmqvist <fredrik.holmqvist@gmail.com>
* added optional feature package for libpng 1.5.12 gcc4/gcc2 x86 and ppc
* drop libpng sources and headers from the tree.
* added optional feature package for jpeg 8d gcc4/gcc2 x86 and ppc
* drop jpeg sources and headers from the tree.
* In cache_abort_sub_transaction(), the original_data can already be freed
when the block is being removed from the transaction.
* block_cache::_GetUnusedBlock() no longer frees original/parent data - it
now requires them to be freed already (it makes no sense to have them still
around at this point).
* AFAICT the previous version did not have any negative consequences besides
freeing the original data late.
* cache_abort_sub_transaction() was setting the transaction_next pointer to
NULL in order to remove a block from a transaction -- however, it forgot to
actually remove it from the transaction's block list.
* Minor restructuring.
All controls were enabled when toggling fullscreen on and off;
now it recalls the argument last used in SetEnabled() and reuse it
when called in AttachedInWindow() (which occurs in such toggling)
Fixes#8804.
IconView is now only setting the icon with BAppFileInfo when the file is an application,
otherwise relies on BNodeInfo. This avoids adding ressources within data files.
The IconView also wasn't populated correctly when used on data files (by opposition to
application binaries). It now loads the same icon that is shown in Tracker.
* This fixes a regression I introduced in hrev43648; the mouse cursor
was only being hidden in case the controls were visible (ie. only
after an extra mouse move in full screen).
Renamed {32,64}/int.cpp to {32,64}/descriptors.cpp, which now contain
functions for GDT and TSS setup that were previously in arch_cpu.cpp,
as well as the IDT setup code. These get called from the init functions
in arch_cpu.cpp, rather than having a bunch of ifdef'd chunks of code
for 32/64.
Hoard reserves a chunk of the address space to grow the heap into.
As there is a much larger address space available on 64-bit systems,
we may as well reserve a larger chunk of address space (64GB for now,
though could probably reserve a lot more than that and still leave
plenty of room for other areas).
* Selected bg uses B_MENU_SELECTED_BACKGROUND_COLOR
* Selected text uses B_MENU_SELECTED_ITEM_TEXT_COLOR
* Unselected text uses B_MENU_ITEM_TEXT_COLOR
Update BStringItem, but also the custom Listitem code in the
Appearance and Locale preflets.
Before its name was a lie, since nothing was cached.
Another boolean was added because getting the localized name could fail, and we
don't want to pointlessly try again, so relying on fHasLocalizedName won't work
for that.
Since in my tests this was getting called up to 4 times per application when
opening the Deskbar Application menu, this caching should speed that up a bit,
at least when this file name translation feature is turned on.
Now that we use the actual selected menu item ui_color, this tinting is not
needed. In fact it makes the selected item too dark.
Thanks diver for noticing.
Omitting this call left invalid reference to deleted group in the
multicast groups hash-map and any upcoming attempt to join the group will
crash the system. Fixes#6736.
- If a trace entry has a stack trace, attempt to demangle the associated symbols.
Could be enhanced further to also demangle the arguments but doesn't yet.
Interestingly there are some mangled symbols that our demangler appears to
not handle correctly (gcc4).
Double-click check was redoing what is essentially already done in input_server.
The way we were doing it, right clicking (or pressing a different button for the second click,
for that matter) wasn't clearing the fields remembered and thus not breaking the sequence.
So a third click returning to the correct sequence (in a short time) would get recognized
as a valid second click. So a quick left-right-left would be seen by that method as left-left.
Also, clean up a previous fix I committed. Removed the introduced Origin() method as it
is the LeftTop() method I was interested in and it is already existing.
Fix#8714
Kernel mode code on x86_64 needs to be built with -mno-red-zone as
interrupts would corrupt the red zone if it were in use. However, the
kernel is linked with libsupc++, which was not compiled with
-mno-red-zone. If an interrupt occurred in libsupc++ code the red zone
would get corrupted. This was causing random panics, particularly under
heavy system load. Therefore, on x86_64 a separate build of libsupc++
with -mno-red-zone is now done for the kernel to use. Note: this commit
will require a rerun of configure and rebuild of cross tools.
This also matches the client_window_info.show_hide_level field used in Deskbar
and other applications.
While doing this, keep fShowLevel fully in sync between BWindow and app_server,
use one message type for both hiding and showing, and make the decision to show
and hide the window in the app_server.
Lastly make minimize behave as described in the Be Book: hidden windows cannot
be minimized, and minimized windows which get hidden become unminimized.
Initializing the IO-APIC will initialize the PCI module, which does
read the MSI config of the devices only when MSIs are available.
Since we initialized them only after that, that condition wasn't met.
Later, due to the uninitialized arch info, MSIs were still marked as
available (0xcc = 204 MSIs). Due to the also uninitialized configured
count, they were always deemed busy however, in effect just breaking
MSI support whereever IO-APICs were available.
Replaced remaining "Preferences" and "Options" with "Settings" as
that is generally used for app settings instead of the system
preferences found in the preference panels.
Renamed Tracker's "Preferences" to "Tracker preferences" to be
similar to the entries in the Deskbar and e.g. the Media replicant
in the Deskbar tray.
TextWidget tried to detect if the editing box would span outside the PoseView,
but it was using an hardcoded value of 1 for the minimum left value. But in Icon mode,
negative values can occur. Change to use the view's origin (top left corner of the view).
This reverts commit 14b654326d.
Unfortunately that changeset causes a regression on GCC 2, which
makes playback of (some?) video impossible. This is due to Libavcodec
being miscompiled, which requires gcc >= 4.2
Resolves the regression of #8856, but does not fix the root issue.
Due to clipping of a rect to match the view bounds, there was a confusion
as to whether the rect was at the top of the view bounds, or above the view
bounds as both met the condition.
Fixes#8876.
This bug was introduced by changing IS_USER_ADDRESS to check against
USER_BASE AND USER_TOP rather than just !IS_KERNEL_ADDRESS. Faults
on addresses outside both the user and kernel address spaces (i.e. the
gap between user and kernel) would result in addressSpace being NULL,
but addressSpace was being used without checking for NULL at one point.
If an uncanonical address is accessed a general protection fault will
be raised. When in the debugger, uncanonical address faults should be
handled by the fault handler (if any).
The coreutils sources get built against the regex.h included with
coreutils, but linked against the regex implementation in libroot. Some
of the types used by coreutils' regex implementation differ from those
in libroot's on 64-bit platforms meaning they are incompatible. Fixed by
building coreutils' own regex implementation rather than linking to
libroot's (GCC 2 still uses libroot though, coreutils' implementation
doesn't build with GCC 2).
This reverts commit df62aa28a7.
Nope, this did not work, it added a new file instead of renaming
the existing one, so there are 2 French (Bépo).keymap files instead
of one, tricky.
* fs_darwin.c => fs_darwin.cpp
* fs_freebsd.c => fs_freebsd.cpp
* use bool instead of int again in fs_darwin.cpp (C => C++)
* declare loop varibles inline again in fs_freebsd.cpp (C => C++)
* 2 newlines between top header gaurd and first #include
* 2 newlines after last #include
* freebsd/endian.h and freebsd/regex.h convert \r\n to just \n
* remove some leading tabs in fs_freebsd.cpp
* add newlines after single line if statement in fs_freebsd.cpp
* 80-char limit fixes in fs_freebsd.cpp
Implementing the window_info.show_hide_level in terms of this solves the
problem of minimized windows also being considered hidden, when really they are
just hidden in the app_server.
window_info.show_hide_level is still defined backwards with a comment making
that clear.
Also removed sending fShowLevel in the minimize request since it is now
maintained in the app_server.
Fixes#4127.
This completes the final 1/3 of #8857. Changes again by nielx with
style fixes by me.
The one part that I couldn't figure out, and maybe Ingo can chime
in here. If headers/build/host/darwin/sys/stat.h is surrounded in
extern "C" {
}
guards then I get a link error complaining that the functions defined
here are duplicate symbols, once in fs.o and once in function_remapper.o.
For example:
ld: duplicate symbol _futimens in generated/objects/darwin/x86_64/release/build/libroot/libroot_build_function_remapper.a(function_remapper.o) and generated/objects/darwin/x86_64/release/build/libroot/libroot_build.a(fs.o) for architecture x86_64
I'm not sure why that is.
- When an Identify/Force Identify request is made in Tracker, if the target
is a link, resolve it to its destination first. Fixes#8858.
- Have mime_update.sh explicitly mimeset the welcome/user guide scripts.
This avoids selecting those windows for moving and more importantly bringing to
front a minimized window with Ctrl-click, which causes the minimized state to
be out of sync between the app_server and BWindow, causing #8860.
Since the position of the widget was registered at the first click,
it likely changed and its causing drawing afterfacts (it's editing at
its old location).
Make the PoseView stop watching a TextWidget if it's being deleted.
Could happen in race conditions for example, if you click to edit
the name widget of a pose while the pose is being deleted soon after.
* Various compilation fixes.
* Fixes to the FreeBSD compatibility layer (from comparing the x86-
specific bits with the equivalent amd64 sources in FreeBSD).
* Compile all the Ethernet drivers except for sis900 and wb840, these
require a bit more work to fix (will file a ticket soon). Tested
ipro1000 and rtl81xx, no issues.
Some preference apps, mount_server and AboutSystem. Removed the check
for x86_64 in the boot script, the normal path through the script will
work now. Also removed a temporary hack to workaround AboutSystem not
being there in build_haiku_image.
Don't wait for a potential second click (and then trigger Widget editing) when:
1. a click occurs on a different pose, on a 'pose-less' area or when right clicking
2. when you start dragging something.
Make the "second click of a double-click" detection waiting time async. In other words
(hopefully clearer), when the TextWidget gets a click, it register itself, recording the time,
and it will get the editing order later as a callback from PoseView when the delay without any
further click expires.
Fixes#8818 and maybe others.
Now that NFSv4 almost works, add support for handling nfs: urls to urlwrapper.
Should later be replaced by a better solution but it works for the time being.
The type of BRegion::fCount is long, and ServerLink sends/receives it
with sizeof(long), but LinkReceiver was using sizeof(int32). Due to
long being 64 bit this was resulting in a mismatch. This fixes the
drawing problems on x86_64.
Previously, changing a file's icon was creating both an attribute
and a ressource, now it only saves it as an attribute
i.e. don't change the file content.
This became visible when dragging an image together with a cursor that has a reasonable large hotspot. In this case the cursor and the bitmap were shifted to much.
In rare cases such as described in #7955 BTextView happens to calculate the
width of a tab close to zero (e.g. 0.000031). This patch adds a fallback to the
default tab width in that case.
Signed-off-by: Ryan Leavengood <leavengood@gmail.com>
Due to a previous change in the incrementation of models in AddPosesTask,
PoseView containing large number of poses were prone to crash in ClearPoses().
Gaps in PoseList or other invalid memory area were introduced.
Thanks to x-ist for precious assistance in spotting those issues!
Again, a regression fix. Under some circumstance, fFiltering was wrongly
set. Also, the node wasn't open before checking the BRefFilter and some
filters expected that to be.
Regression of hrev44493, file panel crashed when changing dir and
filtering being disabled (sorry!)
Also fix the widget editing when filtering is activated. Now renaming
a file/folder and causing it to 'fall off' the filter does work.
Following hrev44493, the way ShouldShowPose was used changed a bit,
to harmonize with type ahead filtering.
This updates OpenWithWindow to use a BRefFilter to do its
filtering rather than overloading ShouldShowPose.
1. A BRefFilter-ed PoseView wouldn't keep monitoring files after its opening
thus not picking up files that happen to fit the filter thereafter (mime attr
updated, etc..)
2. A filtered PoseView wouldn't get updated when:
a) a column was added or removed
b) a file was renamed
c) a file was moved
3. Harmonize the way BRefFiltering and Type-ahead filtering are working. Both
can be used together.
4. The handler for AttributeChanged() wasn't working properly if link(s) of a
changed model was/were its siblings.
5. Broken links weren't detected/updated (it nows monitor the lost target
parent directory, and wait for the target creation to show the link fixed)
This is a big change (even more considering the 'critical-ness' of the component)
Testing is MUCH welcome!
Should fix#4254, #5381, #1717 (and maybe others)
As mentioned in one of the previous commits, breakpoints don't work
properly yet, and I haven't done much extensive testing yet, but the
basic functionality works.
Reused x86 arch_user_debugger.cpp, with a few minor changes to make
the code work for both 32 and 64 bit. Something isn't quite working
right, if a breakpoint is hit the kernel will hang. Other than that
everything appears to work correctly.
* Force fractional feedback divider on APU's
* Spread Spectrum is now probed more correctly
across multiple encoders and cards
* SS still disabled however.
* Move DisplayPort external pixel clock out of pll
as this frequency is card-wide.
* Add new function to pull display clock frequency
and other card-wide settings.
* Set displayDefault frequency card-wide
* My DisplayPort LVDS bridge laptop now kind of works
(a clock somewhere seems a little off though)
There were two problems with the last commit:
* the list needed to be outside of the top-level loop.
* BList was just broken for sorting translation_format pointers.
I fixed this by moving the loop outside and converting the translation_formats
to translator_info, which has the translator_id, since that is needed to create
the menu item, and would otherwise be unavailable outside the loop.
I tried to get this working with BList, but the sorting was completely broken,
and converting to BObjectList made the code much, much better and worked great.
Screw BList and casting, hurray templated BObjectList.
Really fixes#6782.
* Commented out by default
* Shows DisplayPort status info for each
connector post-mode change (as DP properies
are configured on mode change once we know
the pixel clock)
Not in duplicate fragments, however. This oversight in hrev44359 made
checkfs incorrectly rebuild valid indices, causing #8762.
Signed-off-by: Axel Dörfler <axeld@pinc-software.de>
Fixes were needed to go with the USB stack changes. Haven't tested it
yet because I don't have a a machine with OHCI, but I doubt there
will be any problems.
- Various changes to printf format specifiers to make them 32/64 safe.
- Adjustments to various descriptor structures since addr_t is not 32-bit
on x86-64, and consequently this breaks the physical structure that's
expected. Also adjusted padding to accomodate 32-byte boundaries when
64-bit pointers are in use.
- Adjustments to make use of phys_addr_t where needed.
This is used by ShowImage and CodyCam to create a list of image formats which a
file can be saved as. Tracker sorts the image MIME types used in the Find
window by name, so this makes these Save As menus match that (minus the icons
which I think are superfluous.)
Fixes#6782.
If the use of BList is no longer recommended, I welcome better suggestions
for sorting which will work in both GCC2 and GCC4. But this works ;)
Should not be a functional change. It is not in the Haiku Coding Guidelines but
I feel like 'if (object != NULL)' is generally preferred to 'if (object)', plus
in this case of be_control_look that is the more common style.
Added SetFlags(B_CLOSE_ON_ESCAPE) or SetShortcut(index, B_ESCAPE) to BAlerts
depending if the result gets used later in the code, or if it's a one-button
BAlert.
* Remove non-generic radeon dp_get_lane_count
* Set lane count and link rate at set_display_mode
* Pass entire mode to pll_set vs only pixel clock for DP code
* Add helpers for DP config data to common code
* Obtain more correct link rate
With this commit, app_server now compiles and runs at boot! Nothing
particularly interesting happens, just the blue background and a mouse
pointer. Remote backends are broken and not compiled in, see #8834.
Note that it won't be possible to build this quite yet, need to get
the FreeType package uploaded.
* "thread": prints/sets the current thread.
* "continue", "stop": continue/stop the current thread.
* "sc"/"bt": Print a stack trace for the current thread. Very basic yet.
* Introduce the notion of a current thread. That's the one certain
commands will target (by default).
* Add more event handling in CliContext. There's now a queue of pending
events which are printed in the input loop at convenient times to
inform the user about what happened (new/deleted/stopped threads,
etc.).
An off-by-one error (> instead of >= really) was causing truncation when it
wasn't needed. But for some reason this only showed up when the main font size
was increased. I may or may not look into that, but this fixes the issue.
Maybe one of the faster bug logged to bug fixed times in Haiku history? ;)
* Make sure we use consistant terminology
* No more Formatted partition "" messages when
doing a partition map
* This commit will require translation work
No functional change.
* Surround email addresses in angle brackets.
* Add myself to ExpressionParser.cpp and .h
* Remove myself from ExpressionTextView.cpp and .h
* Alphatetize authors by last name.
Thanks Ingo and Axel.
This module provides an interface for drivers to use to perform calls
to the BIOS (only really for use by graphics drivers which need to use
the VESA BIOS). It uses the x86emu library from X.org which emulates
a real mode x86 CPU. This is necessary for x86_64 as virtual 8086 mode
no longer exists there.
* Move disk partitioning systems under Disk menu
* Found and created bug #8827 (this wasn't introduced
by my changes and was pre-existing)
* My testing resulted in no regressions, however
please test though!
Figured out by Stephan originally and only hacked in WebPositive, I knew
something about WebPositive was different because it did not flicker when
everything else did.
Since we fully draw the menus there is no need to have app_server fill in the
background with the view color. That is what was causing the flickering.
Should fix#484, #532 and #4335. Thanks diver for confirming the patch against
those tickets.
* Changed IS_USER_ADDRESS to check an address using USER_BASE and
USER_SIZE, rather than just !IS_KERNEL_ADDRESS. The old check would
allow user buffers to point into the physical memory map area.
* Added an unmapped hole at the end of the bottom half of the address
space which catches buffers that cross into the uncanonical address
region. This also removes the need to check for uncanonical return
addresses in the syscall handler, it is no longer possible for the
return address to be uncanonical under normal circumstances. All
cases in which the return address might be changed by the kernel
are still handled via the IRET path.
* Delete and q quit the application.
* Shift-Delete and k kill the application.
Used a message filter since I want these handled by the window.
Fixes#8561.
Also:
* Removed DeselectAll() from the class declaration since it was never defined.
* Re-ordered methods in the class definition to match the order in the declaration.
* Added the underscore to _UpdateList() since it is a private method and moved
it to the right place.
The ellipsis character (...) after a menu item means that
the command needs more information from the user before
the operation executes.
The ellipsis character doesn’t simply mean that a dialog
box or window will appear.
The window appearing completes the command and doesn’t
require additional input from the user before it executes.
The current behaviour of aligning to the maximum value of p_align seen
is problematic for x86_64, as the default segment alignment is 2MB.
This causes all x86_64 binaries to be padded to at least 2MB when
resources are added to them. There is no need to align to p_align in
the file itself (it's only an in-memory requirement), therefore
instead just align up to an 8-byte boundary. The current behaviour is
retained for ELF32, so this won't cause any compatibility problems
(there are no existing ELF64 BeOS/Haiku binaries to worry about).
I overlooked this problem in my last commit...
If you select the current angle (radian/degree) mode from the menu
it toggles the angle mode, fixed this to make it keep the current
mode.
Default is radian mode, You set the option in the right click menu
like the other options.
Note: degree mode does not affect hyperbolic trigonometric functions.
This is how Mac Calculator, Windows Calculator, and Google Calculator
work.
I'm using consoled at the moment without input_server having been run,
so when it starts the input drivers will not have been loaded. Have to
retry a few times with delays so that the input devices have a chance
to appear.
Fixed the usual issues - printf format strings, uint32 instead of
addr_t, etc. One thing that isn't so nice is several places where
BList is used to store (u)int32, these require a double cast to addr_t
then void* to silence a warning on x86_64.
This reduces perceived latency. One day when I add a tab close undo stack, this
will be even better. But generally using the middle mouse button is not
accidental.
The USER_BASE_ANY definition exists to specify where to start searching
for B_ANY_ADDRESS allocations, but this was not being used correctly.
On x86_64, this was causing the runtime loader's heap to be allocated
at address 0 so NULL pointer accesses were not getting caught.
- When loading an image, we now check if it has a .gnu_debuglink section.
If so, we try to load the linked file for our symbolic debugging information.
The file is searched for with the following rules:
1) If self-contained app, app image is looked for in ./debug while
library/add-on images are looked for in ../debug/.
2) If the image in question is in one of the global locations
(/boot/{common,system,home/config}/{lib,add-ons}), we search
in develop/debug for the corresponding global dir.
Implements #8135.
The cookie is used to store the base address of the area that was just
visited. On 64-bit systems, int32 is not sufficient. Therefore, changed
to ssize_t which retains compatibility on x86 while expanding to a
sufficient size on x86_64.
Refactor the icon scaling code in IconUtils.cpp to avoid code
duplication. Basically create and delete the temp bitmap to
convert from B_CMAP8 to B_RGBA32 for scale2x/scale3x/scale4x
just one time instead of 3.
There was an off-by-one error in Deskbar which was causing
it to scale up the 16x16 Bitmap icon to 32x32 instead of just
using the 32x32 icon. This only affected BeOS bitmap-based
icons, not Haiku HVIF icons.
* The input loop can now wait on abstract events, which other threads
(or even the input loop thread itself) can signal.
* Use the new mechanism in QuitSession().
* Also (with the exception of the SIGINT part) implement
WaitForThreadOrUser().
* Added x86_64 linker script and relocation code.
* Some 64-bit safety fixes to the heap code.
* Added runtime_loader, libroot and bash to the x86_64 image. The boot
script will be launched, but will panic shortly after because fork
is broken.
These have been pulled from glibc 2.10, the last version before a
bunch of changes were made that would have made porting more difficult.
None of this has been tested yet, it is not currently possible to do
so: I'm just trying to get libroot compiling so that I can work on the
runtime loader. I will test them when I am able to.
* Move the libedit interface there and provide nicer to use methods.
* Also start adding utility methods for the input loop. It is going to
manage all interactions of the input loop with outside events.
* Fix the "quit" command. The user is now prompted what to do with the
debugged team and the input loop thread avoids reentering the input
loop.
* UserInterface::SynchronouslyAskUser() is now allowed to return -1 to
indicate that the user cannot be asked at this point for whatever
reason. The caller needs to handle that case.
* UserInterfaceListener::UserInterfaceQuitRequested(): Add new parameter
"quitOption" to specify what is supposed to happen. The previous
behavior (ask user) is only one of the options. The others are to kill
the debugged team or to resume it.
* Create a new Haiku GPT GUID (BeOS type not defined atm)
* Haiku BFS UUID by Andre Braga circa 2009 ML post
"Defining the Haiku UUID for GPT and other uses"
* I'm putting this GUID on wikipedia and pushing to
the linux gpt partition tools... should be a good
way to kickstart it in the ecosystem
The interrupt and system call handlers now perform all the necessary
kernel entry/exit work, and the system call handler now handles calls
with more than 6 arguments. Debugging and system call tracing hooks
are not yet called, will be added when user debugging gets implemented.
- When invoking ProcessController's menu, we now only show the "Live in Deskbar"
menu item if we're either running within Deskbar itself or from PC's standalone
window. This allows replicant PC instances to be usable in the case where
Deskbar is deadlocked for whatever reason (previously it would hang while trying
to query for the deskbar item's presence/status).
Right clicking on a Pose to get the contextual menu would quite often
trigger a rename action of that pose. Don't allow to rename a pose
by releasing the secondary mouse button.
When restored, an overlap was wrongly detected in offsets for
failure to take into account the width of the border line.
This was causing the horizontal scrollbar to show unnecessarily.
A crash of Tracker was triggered when accessing AddOn menu (by
shortcut or context-menu) for Pose on Desktop, because of it's
incapacity to read the mime type list (that wasn't built in
those cases).
* You can code review it
* You can help developping
Uses libvterm as the backend for parsing ANSI escape sequences. The lib was
changed slightly to build with GCC2. It could be used by Terminal as well as
it seems cleaner and more reliable than our current parser.
1. Build the list of mimetypes of files in selection only once and
reuse it for all further tests.
2. Fix a regression introduced in hrev44384 where the MimeType()
wouldn't get recognized when just changed by tracker (by that same
right click). It would be on subsequent clicks.
3. Rename the static map variable to better fit our coding style
and be more understandable.
Userland switch is implemented, as is basic system call support (using
SYSCALL/SYSRET). The system call handler is not yet complete: it doesn't
handle more than 6 arguments, and does not perform all the necessary kernel
entry/exit work (neither does the interrupt handler). However, this is
sufficient for runtime_loader to start and print some debug output.
When sorting files by Modified dates, right clicking on a file was leading
to a sorting issue where files were changing positions (without reason).
1. Any changes to stats (size, modification, creation, mode) was triggering
the sorting. Now only stats fields currently used as a Sort criteria will
trigger such event.
2. The Mimeset of file was set (in case of unknown file format) once per checked
add-on when building AddOn Menu. Now it's checked once per file in selection.
(so, once per file, rather then once per file, per add-on).
3. Now rely on registrar to force the mimeset (to trigger the sniffer in case
the attribute already exist) rather than trying to duplicate the feature in Tracker.
4. When Sorting, if there is a old position known, check if it's working by looking
if you should come after the previous item, and before the following item. Previously,
the item would be pushed at the top if the group of item all fitting the criteria
(same file size, same file kind, etc.. depending on the sorting criteria).
Fixes#8478.
- When the message filter would receive and process a mouse moved message,
if the message resulted in causing the deskbar to relocate or reorient
itself, it was possible for the expando view to become detached from
the looper. Consequently, if the intercepted mouse moved happened to have
come from the latter, when returning out of the filter the view would no
longer have a target looper, triggering a debugger condition in BLooper.
In order to prevent this situation, we now dispatch a message asking for
the layout change to occur asynchronously.
* Option devices are generally WWAN serial
devices for 3G or lower.
* Picks up my CMOTECH Sprint 3G adaptor, need
to wire up endpoints so disabled for now.
Since the demangle debugger extension now gets loaded when booting
from an image, use it in stack traces. Can't print argument values
like on x86, however, since x86_64 uses registers to pass the first
6 arguments rather than the stack we can't easily get to them.
Added a temporary Haiku64Image file that gets included instead of
HaikuImage when building for x86_64, which I will add to as I port
stuff. Images currently only include the boot loader, kernel and
a bunch of add-ons.
Fixing #1592.
A feedback FSNotification()->EntryMoved->PendingNodeMonitorCache->FSNotification
was seemingly introducing some race condition, as it was working 1 time
on several tries.
Since the commpage is at a kernel address, changed 64-bit paging code
to match x86's behaviour of allowing user-accessible mappings to be
created in the kernel portion of the address space. This is also
required by some drivers.
Since this argument may be used to pass pointers, uint32 is not
correct for 64-bit. Effectively no change on 32-bit targets, both
size_t and uint32 are unsigned long there.
Using "kernel.h" was pulling in the private kernel.h header instead,
which was causing a build failure on my branch since arch_cpu.h is
C++-only there.
* Update FTDI, KLSI, Prolific, and Silicon drivers to share a
common structural layout for device identification.
* More flexible and cleaner than massive switch case statements.
* Avoids the problem of different chipsets from identical vendors.
- Store whether or not the use of the horizontal scrollbar is desired
on the class itself. If the CLV was set to use the horizontal scrollbar,
and then asked to lay itself out while hidden, it would incorrectly assume
the horizontal scrollbar wasn't in use, and consequently repositioned its
views such that the horizontal scrollbar and outline view overlapped.
There's an input loop thread which reads and parses command lines and
the infrastructure for registering and executing commands. Currently
only "help" and "quit" commands are implemented.
Mostly compilation fixes, as well as a few 64-bit safety fixes. I've
briefly looked through everything for any obvious issues and fixed
the ones I've found, and it seems like they're working properly, though
there could be some more well hidden ones that I've missed.
Uses the x86 architecture code, made fixes to printf formats and a
couple of 64-bit fixes. Only potentially intrusive change is that I've
changed PCI.h to use uint32 rather than ulong. I don't see any way
this would cause any issues, though.
- When a job needs to go dependent on another job, we no longer use recursion
to manage the execution stack. Instead the job is simply marked as waiting
and we execute other jobs with no dependencies in the meantime. When a job
completes, all dependents are moved back onto the unscheduled list and
executed as needed.
- Adjustments to ResolveValueNodeJob to handle the now asynchronous nature
of waiting.
Such arrays could be created by the BFS code between hrev43837 and
hrev43924, and cause the array to not be free'd when the entry is
removed.
The check in _InsertDuplicate() is not changed, as doing an insertion
will actually repair this problem.
Potentially, the code could be modified to handle this state instead,
but since checkfs can fix it, it doesn't seem necessary.
Signed-off-by: Axel Dörfler <axeld@pinc-software.de>
* cache_abort_transaction() left the block dirty which was causing bug
#8123 as well.
* cache_abort_sub_transaction() did, in addition to not clearing the dirty
flag, not reset the block's transaction member either if the block was
not part of the parent transaction.
* add a sniff rule
* until we have an image translator for SVG, and since Web+ is not
included by default in Haiku images, made Web+ preferred app for SVG images.
Note that for compressed SVGZ (*.svgz), the sniff rule will still fail miserably.
Our sniffer will need decompression support before being able to do that...
Added the necessary build flags for modules, and added a module (dpc)
to the floppy image for x86_64 builds for testing purposes. The module
gets loaded correctly and its code runs without issue. Only non-trivial
addition is the different method for generating kernel.so, this is
explained in the kernel Jamfile.
- If the node we're looking at is a compound node, retrieve its location
and show that as the value with an indicator to clarify that it's an object.
- Minor tweaks to tooltip format.
- Style cleanups.
Our FreeBSD networking code defined hz to 1MHz and 1 tick = 1 / hz,
but the clock code ticked 1 tick at 1000Hz.
This caused all calculations that are done on ticks, autonegotiation
and wlan scanning to be done very often as FreeBSD uses 1000 Hz
(100Hz for ARM). Defaults for autonegotiation is 5 and 17 ticks.
(Another interesting thing is that callouts are using 8% cpu...)
- When resolving its value, CStringValueNode now sets its node child's
address to the address of the string buffer rather than the location of
the originating pointer, which allows things like Inspect to pick that up.
- TableCellContextMenuTracker now supports menus that don't have a settings
submenu, since some variables won't have renderer settings but will still
have context actions.
- Add _GetContextActionsForNode() to retrieve the list of contextual actions
available for a given model node. Currently this is only adds an action
to inspect the memory address of the highlighted value, but will be extended
for other actions later.
- If we're asked to generate an entry for a tag we don't recognize,
return an error so don't then attempt to add a null/invalid entry into the
compilation unit's entry list and later crash dereferencing it.
- When asking for signed/unsigned, IntegerFormatter was previously
treating all value types as int64, which would result in much larger
than expected values in the variable list in some cases. Inspect the actual
integer type of the variable and adjust the format string accordingly in
order to deal with that.
- If a string column was exactly the correct pixel width for the strings
contained within it, they would skip attempting to truncate the string.
However, the truncated string was always used for final drawing, with
the end result that the fields would be drawn blank. This would sometimes
manifest itself in Debugger where things like ID columns wound wind up
showing no data until one resized the column.
* Patch from #6885: This filters out any disallowed characters from text
being pasted or dropped in. If the resulting filtered text is zero
characters long, it beeps. Works with styled text too.
* Coding style: variables renaming by korli.
This is an update from 20110922 and A LOT has happened since then. See
https://acpica.org/download/changes.txt for all the changes.
One of the more interesting is some problems with shutdown in ACPICAhas been fixed.
There may still be problems on our side though.
Not sure why but GCC 2 seems to be ignoring the extern "C" in smp.h when
compiling smp_boot_other_cpus() resulting in undefined references. Fixed
by putting extern "C" on the definition too.
Currently all debugger commands assume 32-bit pointers when formatting their
output. This means that on x86_64 the output is incorrectly formatted. Fixed
this by adding a B_PRINTF_POINTER_WIDTH definition (16 on 64-bit, 8 on
32-bit), and using this to correctly format the output. Not all commands have
been fixed yet, but all VM, slab, VFS, team, thread and image commands should
be correct.
- When binary searching functions in the source entry list,
comparing by name and location alone isn't sufficient, since
templates will match those for different instances, Fixes a crash on
exit where the wrong function would get removed from the list, while
the one we actually wanted to remove was still in the list, but then
had its source code cleared. This would later crash the comparison
function due to not being able to get its source location.
- When an image was unloaded, its corresponding image info was never
removed from TeamDebugInfo's list, leading to the latter containing
a deleted object, resulting in various random crashes.
- When attempting to unwind the call frame, we now search for the appropriate
FDE in both .debug_frame and .eh_frame. This mirrors gdb's behavior and
works around the ever-changing whims of the gcc developers as to which
section the requisite FDE/CIE resides in.
