This didn't cause any problems before because all the views that don't
use layouts reimplemented this function. But that may not be the case in
the future.
He's doing such a good job promoting Haiku as well as maintaining
our (extensive) lists of articles on the website, so there's no reason
not to credit him here.
Intel changed the PCH interrupt bits between Sandy Bridge and Ivy Bridge
to make space for the 3rd display pipe. Take this into account and check
for the correct bits on the newer devices.
Fixes#11522.
This reverts commit de7aed2b56. It turns out
that this causes crashes when trying to navigate in directories with
type-ahead filtering enabled. (possibly related to the root cause of
Rename CopySelectionListToBListAsEntryRefs to CopySelectionListToEntryRefList
to reflect that it no longer copies to a BList but instead to a BObjectList<entry_ref>
When items were removed from fFilteredPoseList as opposed to fPoseList, the
items were not automatically deleted and thus leaked. By setting the BObjectList
to automatically delete them on removal (by setting the deleteIfOwned paramete
to true) we avoid this leak.
fWidgetList is a BObjectList, the items are deleted when removed already so we
should not have the BObjectList delete them for us too. Instead, let's just turn
that off so we can delete the items ourselves one time.
All the people I've moved off this list have not committed anything
since (at least) before 2012-01-01, with the exception of Andrew Lindesay
who never recieved commit access (AFAICT, he only contributed some patches.)
All the packages I delete from here are not included in the tree
and not included in the "default packages" list.
* Use HTTPS rather than HTTP for sites that support it
* GDB was removed in hrev48791
* FFmpeg is spelled as "FFmpeg", not "FFMpeg"
* atftp was removed from the tree in the previous commit
* acipca was updated to a version from 2014 recently
* CannaIM was removed in hrev48034 & hrev48035
* libxml2 & friends were removed in hrev48798
* Cleaned up ifdef __INTEL__ lines -- no need to have more than one
If you have no apps open because Tracker crashed and you opened a new team it
would size the expando menu bar incorrectly in vertical mode.
This fixes the problem.
Let me tell you a story about a bug.
In a615ec8f86 Axel imported the OpenTracker code
into Haiku with fSaveStateIsEnabled set to true (for Desktop and all windows).
In 9f2a3f6f47 I (John) changed it to:
fSaveStateIsEnabled(dynamic_cast<BDeskWindow*>(this) == NULL) which is suppose
to detect whether or not the ContainerWindow is the Desktop and only save the
state if it is NOT. However, this didn't actually work because since this is
called in the constructor the dynamic_cast to BDeskWindow* fails. Therefore we
were still saving the state of Desktop even though the code make it look like
we weren't.
In e7803cf1f6 Augustin updated the code so that
BDeskWindow passes in a isDeskbar flag as part of the constructor parameters
thus detecting that the window was Deskbar correctly. So basically Augustin
fixed the bug detecting whether or not the window was the Deskbar in the
constructor which then disabled saving the Desktop's state to attributes.
Augustin unknowingly turned saving the Desktop's state off.
We don't want to disable saving the Desktop's state because that's where we
store things like the icon size. What I was probably thinking when I tried
unsuccessfully to turn off saving the state of the Desktop was that I don't
want to save the WINDOW state of the Desktop since that would save the
window dimensions of the Desktop (whatever your screen resolution is) and if
you changed Tracker to non-spatial mode and opened the Desktop in a Tracker
window it would pick up those attributes and then make the window take up the
entire dimensions of your screen (a whole different bug).
This bug, however, is also handled in 9f2a3f6f47
because code was added to BContainerWindow::SaveWindowState() which detects
whether or not the window is the Desktop and if so, doesn't save (or restore)
the window state (which is only part of the state).
So with this change (reversal really) the other parts of the Desktop's state
like icon size are once again saved while the window state is not.
Fixes#11864
Replace calls to dynamic_cast<BDeskWindow*>(this) != NULL with fIsDesktop
which is the flag added to the constructor in
e7803cf1f6.
We no longer need to do the expensive dynamic_cast's because we have a handy
flag to check instead.
We must toggle the expanded state of new windows if we are in
VERTICAL expando mode with Expand New Teams on.
The vertical part of the check got dropped accidentally in hrev48817
Thanks jstressman for reporting.
Fixes#11865
This is all work in progress. The plan is to move the code that does any
actual changes to the TextDocument into UndoableEdit implementations, then
emit these edits to interesed listeners. They can then store them to
implement the edit history. If there are not listeners, the edits will
simply be released after they've done their work.
Implemented so far is only the support for storing the edit listeners.
