- It is possible to call open() on a directory, but FUSE lowlevel
filesystems don't implement that and expect it to be re-routed to the
opendir call. BRoster uses this to read the dir/file attributes to
identify it, so it could not identify directories properly.
- In ReadDir, make sure to not return more entries than asked, as this
confuses the userlandfs protocol communication (the kernel does not
acknowledge the readdir reply, and then the server hits an assert when
receiving the next request instead of the ack).
Change-Id: I9c4e9a3f0fc6e9879d4cfbc0d5402a4733d2218a
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5482
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
The previous implementation needed a temporary buffer to store entries
and then sent them to the actual readdir buffer. This is now fixed, the
actual buffer is filled directly.
Also fixes problems with reading the same directory multiple times, and
reading large directories.
Change-Id: I8dc9677ee676144547d17f313a7f2d91fd2bca05
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5480
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
- Import latest version of files from FUSE 2.9.9 (our last
synchronization was with 2.7.4)
- Adjust fuse pkgconfig file to use the POSIX error mapper
automatically, since that's required for all FUSE software
- Implement the lowlevel API in addition to the highlevel one. The
lowlevel API uses inode numbers to identify files, rather than paths,
making it a better fit to the userlandfs architecture.
The FUSE 2.x branch is not maintained anymore by FUSE developers,
however, pretty much no one migrated to FUSE 3.x. So it is more
interesting to implement, rather than 3.x.
Confirmed still working with sshfs and curlftpfs.
Example use:
I tested this with github.com/whoozle/android-file-transfer-linux
- Build the fuse library and copy it to ~/config/non-packaged/add-ons/userlandfs/
- Start the server: /system/servers/userlandfs_server aft-mtp-mount
- Connect your Android phone and put it in USB file transfer mode
- Mount the device: mount -t userlandfs -p 'aft-mtp-mount /boot/home/MyPhone -d -o use_ino' ~/MyPhone
- You can now access your phone data
Change-Id: Ic3efda7ffbc33737e6f4958428fb3ec9939ef105
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5198
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
It needs to be deleted after everything is done.
This seems to fix the crashes I could reproduce on exit in my VM.
So might fix#17583.
(WebView also has fContext, but I assume they are part of a window)
Change-Id: I0125f76b0b2331ac3b6a81d6c20087ebdedbf2f2
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5486
Reviewed-by: Fredrik Holmqvist <fredrik.holmqvist@gmail.com>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
... now that Add/FindNodeRef have been added.
Tested by generating thumbnails on read-only BFS file system.
Change-Id: I94d85f4edfb6f26fc88dcb7b85d9d36629062f6b
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5485
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
If you have multiple Haiku installations mounted, it is common to
have the same app on multiple volumes. This makes the Open with...
menu show the full paths truncating them which makes it tough to
see the app name you're looking for.
Show volume and app names for repeated apps in Open with... menu
instead of showing the full path. Keep track of whether each app in
the list is a repeat and only display the volume for those.
If there is only 1 app name, print it.
If >1 app name and volume names are different, print
[volume name] app name
If >1 app name but volume names are the same, print full path
Because many apps are in packages the volume is listed as [system]
Make kMaxMenuWidth a bit wider.
Compare app names and volume names using locale collator.
Sort by name, determine labels, then sort by relation because
repeats are not always consecutive once sorted by relation.
Change-Id: I2c6c1db99065641cf4f940a69a47693480f151aa
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5448
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
This considerably overhauls touchpad event generation, simplifying and
cleaning it up considerably:
* Return the touchpad specifications through the MS_IS_TOUCHPAD ioctl.
* There is now a dedicated MS_READ_TOUCHPAD ioctl, as touchpads
can either return touchpad_movement structures or mouse_movement
ones depending on what mode they are operating in.
* Event repeating on timeouts is now handled in MovementMaker and
the input_server control thread, so MS_READ_TOUCHPAD takes
a timeout value. This means we can drop all the EventProducers.
* Use the real floating-point math functions in MovementMaker now
that we are running in userland.
* Drop unused structures, constants, headers, and other things
related to touchpad support.
Change-Id: I28cdb28e4100393a9338a8ebb865573cec13fc1e
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5455
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
* Change BContainerWindow::fStateNeedsSaving from private to
protected, so that TFilePanel functions can set it to false, like
the analagous BContainerWindow functions do.
* Prevent fStateNeedsSaving from being set to true because of a change
to window size/position, unless the window is active. This is
meant to distinguish changes made during object setup from changes
made by the user.
* Add more calls to ViewState::_StorePrevious state.
BContainer::fPoseView triggers saving on some occasions because its
member fViewState needs to save. fViewState needs saving when there
are mismatches between pairs of f[x] and fPrevious[x], e.g. fIconSize and
fPreviousIconSize. These mismatches can arise when the ViewState
is set up. These pairs are synchonized by _StorePreviousState.
* Rearrange the modified ViewState functions to keep the 'const'
modifier on the existing code.
* Fixes#17374.
Change-Id: I2a6c1c1d9c5c3656eab2ebc770a5915010e7500c
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5447
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
We removed config_manager which the code is based on, and
the driver was not working before that.
Perhaps it is time to pour one out for the floppy driver?
Change-Id: I306f7bda08ad8fe3f64b55cb147a1e2dea6ddfef
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5221
Reviewed-by: Fredrik Holmqvist <fredrik.holmqvist@gmail.com>
* This has been floating around on trac forever. We should formalize
these steps to help future Haiku releases to be successful.
Change-Id: I5881e27a23e66a18539d04c5977593a827f8fcef
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5441
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
- Implemented V5 Block directory fields and virtual extent data header class to handle both V4 and V5 block directories.
- Implemented V5 leaf directory fields and virtual extent leaf header class to handle both V4 and V5 leaf directories.
- Implemented all checks for buffer headers.
Change-Id: I8f994d9ba792f40ba827cd5cee915b7b11796ce3
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5432
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
This code is quite convoluted and very much in need of refactoring,
but this solution works well enough for the time being. It does not
look quite right at higher resolutions (borders come out darker than
they do at the default resolution), but at least the borders are now
the correct size instead of a constant one.
The resize nub drawing code also hasn't yet been updated, so while
it draws the gradient over the full resize area and responds to clicks,
the "checkerboard" pattern is drawn at a fixed small size still.
Nonetheless, this is a significant improvement.
There is nowhere else that unsets it (or makes sense to unset it),
we have to do it here, or otherwise it will never be unset.
This has actually been broken since the new scheduler was merged
many years ago; the old scheduler (which had more complex logic here)
unset this flag correctly.
This oversight mostly did not cause obvious problems: while the scheduler
would be invoked unnecessarily on every single interrupt, that would
be a minor performance downturn at most. But there is one other effect:
since we can only reenable interrupts after invoking the scheduler,
"post_interrupt_callback"s of threads were not invoked most of the time.
The main user of "post_interrupt_callback"s is the profiler, which uses
them to trigger buffer flushes back to userland. Since they were not
invoked, the buffer would quickly fill up and then no more results
would be recorded.
Thus, this fixes#16345.
Extracted as separate patch from CR 4790; implement BFont::BoundingBox
as it was in BeOS. Returns a BRect that encloses any character in a
font, scaled by the font size.
Since the FreeType bbox is only valid for vector outlines, for bitmap
fonts, return a BRect with the dimensions of the bitmap font that is
closest to the font size.
Tested with CharacterMap. Will update CharacterMap to use the bounding
box once this change is landed.
Note that bitmap font rendering doesn't appear to work at all.
Change-ID: I8fc42ca0ddc0c77e04e0216001cd17e086ec7495
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5431
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
The local state scale is applied from the origin. That is, the origin
value does not change with the local scale.
This fixes the value returned by BView::Origin() when recovered from the
app server instead of from the client cache.
Change-Id: Id628d8f334fab06e6a6d084c079cdca1cfa914c6
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5443
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
Implemented the missing POSIX functions in <locale.h>:
newlocale, duplocale, uselocale, and freelocale, and also
provided missing type definitions for <locale.h>.
Implemented missing POSIX locale-based function variants.
Modified LocaleBackend so that it could support thread-local
locales.
Some glibc-like locale-related variables supporting
ctype and printf family of functions have also been updated
to reflect the thread-local variables present in the latest
glibc sources.
As there have been some modifications to global symbols
in libroot, libroot_stubs.c has been regenerated.
Bug: #17168
Change-Id: Ibf296c58c47d42d1d1dfb2ce64042442f2679431
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5351
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
This commit introduces a simple thread-safe ring buffer implementation
based on top of BDataIO. The main use case for this class will be to
implement shared buffers between threads for the upcoming refactoring
of Services Kit.
Change-Id: I526bc044b28c91496ad996fabebe538e75647f2c
Reviewed-on: https://review.haiku-os.org/c/haiku/+/2966
Reviewed-by: Jacob Secunda <secundaja@gmail.com>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
For mouse probing to work reliably, it's important that we don't
simultaneously send commands to the keyboard as well. This normally
happens because we don't publish the keyboard device until the probing
is complete.
However, if the mouse probing results in the keyboard sending some
replies, this can be detected as a "hot plugin" and leads to publishing
the keyboard device a lot earlier.
To prevent this, disable the "hot plugin" feature until we are done
initializing. This way the keyboard device is published only after we
are done with the probing.
Fixes#17806
Change-Id: Ia6a2b031550c845fa305df5b5f4d513d5c7931d6
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5388
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
Reviewed-by: Jérôme Duval <jerome.duval@gmail.com>
Before this patch, calling functions in the `printf` family and
passing %F in the format string, such as
printf("%6.2F\n", 580.0f);
would incorrectly output numbers in the exponential format:
5.8D+02
This patch fixes the output for this situation:
580.00
and fixes the behavior for programs that use %F, such as
several tests for the `fish` shell.
Change-Id: I124a5a7b93b5f551fa190501a5778726812d347c
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5424
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
This is for platforms where information from parsed ACPI namespace
is necessary to fully configure PCI subsystem.
Change-Id: I8bdcfab6b99cbe7fdbc902b9fc13b44133325961
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5273
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>
* Previously, any accelerants failing would cause an
app_server crash.. so we would do everything in our
power to get "something"
* After hrev56252, the app_server properly handles
accelerant failures and vesa / framebuffer accelerants
should will be used.
* Thus, if no monitors were detected (due to none being
plugged-in, or some other fault), bail and return B_ERROR
so vesa fallback can give it a try.
Change-Id: Ib9695dd7cf1914e50547ca8661c55b80152a66a4
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5442
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
Reviewed-by: Alex von Gluck IV <kallisti5@unixzen.com>
