No, not updating packages via HaikuDepot, but a light-weight synchronization
between the system package manager list and the HaikuDepot model. After
installing or removing a package, HaikuDepot performed a very costly update
of all information by basically throwing everything away and reloading it.
Implements #11786.
* Doxygen 1.7.6.1 is not the current version, but it's the last in the
1.7 line which is what we're using for the Haiku Book currently.
* Git 2.2 has some new features (terminal coloring) and behavioral changes
(most appeared as warning messages in Git 1.8). Most major Linux
distros successfully switched, and HaikuPorter doesn't seem to rely
on the old behavior.
* PHP is a new add, I've successfully enabled some of the more commonly
used features in our port (GD, ZLIB, CURL, i18n) so it's more useful now.
This package does not contain any webserver integration files, but it does
contain both the CLI and CGI interfaces as well as the builtin webserver.
* ZSH now provides cmd:sh, so it's possible to switch your default shell to
it. Mksh already provided cmd:sh so I didn't need to modify it. There
are still some rough patches that occur when uninstalling Bash (e.g.
Terminal app has an annoying alert that it can't find Bash so it's using
/bin/sh instead) but it appears to work.
* Go was ported in GSOC 2014 and has had a recipe since then, but nobody
bothered to upload it.
* RCS had an incorrect primary provide, fixed it (#9703).
... in PackageInfoView::SetPackage() and Clear(). The sub-views reference
stuff from the previous PackageInfo instance. If we release the reference
before adopting the new package, the sub-views may still access stuff from
the previous package. For example the PackageActionView tries to avoid
rebuilding the button list and compares previous package actions to the
new actions. Particularily after the package list has been rebuild, we have
only new PackageInfo instances and the PackageInfoView may hold on to a
PackageInfo that is not still referenced anywhere else. Would be a good
explanation for #11785.
BitmapDrawingEngine:
- Check if fBitmap is NULL before releasing its reference. Since this is
the case when a BitmapDrawingEngine is instantiated, this would lead to
an app_server crash upon any attempt to make use of one.
* Use BStringViews instead of BTextView - it's more flexible (truncation,
and tool tips), and allows for fine grained updates.
* The file name and location views are now showing tool tips with their
full contents.
* Also, its look is now less exotic, if boring.
* Added new truncation mode B_NO_TRUNCATION.
* The Truncation()/SetTruncation() methods itself are Dano-compatible,
however, there was no B_NO_TRUNCATION.
LinearSpec is a BReferenceable, so it is not correct to allocate it as a
member of another object. Wrap it in a BReference to avoid the problem.
Fixes #11089.
* BNetworkInterfaceAddress is moved to libnetwork. It is modified to not
use BNetworkAddress (which is in libbnetapi) and instead use sockaddr
and sockaddr_storage directly. All callers are adjusted to this.
* Some support code is shared between BNetworkInterface and
BNetworkInterfaceAddress, move it to libnetwork but in the BPrivate
namespace.
* When the keys were large enough, a large key entering the node could
overflow the available data in the target node. This caused tree and
memory corruption.
* This fixes bug #6034.
The atomic inlines were not implemented in a C89 safe way:
* Use of "static inline" not allowed, but static __inline__ is
* __inline__ is a GCC extension, but these are already in a __GNUC__
block (other compilers use a non-inline version)
* also fix a C++ style comment
* The code was assuming "Attachments" would be the longer word used, but
this doesn't always work outside of english. Actually search for the
longest string in the translations.
* Fixes #5285.
* It would be even better to convert the whole window to use layouts,
but this makes it useable until that's done.
Signed-off-by: Adrien Destugues <pulkomandy@gmail.com>
* Subsequent calls to listen on an already-listening socket can resize
the backlog.
* While not explicitly spelled out by POSIX, this behaviour is
consistent with FreeBSD and Linux.
Signed-off-by: Adrien Destugues <pulkomandy@gmail.com>
* Fixes #11760.
* Bug made visible by the changes to fix private driver settings. Before
the change, the settings were most often leaked, so there was no
use-after-free.
...with notes from PulkoMandy and Axel. Also added author credits.
Class documentation is moved to the appropriate method and then \sa
is used to point to the documentation so it is only documented in
one location.
Added some text about how the interaction between BInvoker and
BHandler and/or BLooper works.
BMessenger needs to be documented to understand how SetTimeout() is
suppose to work, refer to BeBook for now.
Signed-off-by: Adrien Destugues <pulkomandy@gmail.com>
Fixes #8569
I'm not sure if this works properly if you reorder tabs (see #9175). But
the fix for that is reordering the tab list to match the visible order,
which doesn't affect this part of the code.
