* This is needed for redirects to work as expected.
* It makes it harder to send data starting from the middle of a
BPositionIO (you now need a wrapper object), but that is an uncommon
feature so it is acceptable.
Fixes#11687.
As discussed in #11698, selecting an item in the list needs an update of
the selected color, but doesn't need a check for changes and status of
the "revert" button.
This was an old version of bash_completion and not included in the
image. An haikuports recipe is available if you need it.
Fixes#11660.
Signed-off-by: Adrien Destugues <pulkomandy@gmail.com>
The buffer version of Unflatten can be implemented by wrapping the
buffer in a BMemoryIO. This allows sharing more code, and the BDataIO
version of Unflatten has some extra error checks.
Fixes#11670.
* Reorder the languages to match the order used in Pootle, making this
easier to do.
* Remove Diver from hungarian translation on request from him.
* Add some missing languages from Pootle (commented out for now as they
are not exported yet)
SyntaxHighlighter/CLanguageFamilySyntaxHighlighter/SourceView:
- TeamTypeInformation is now passed as a parameter when requesting parsing,
so highlighters can make use of it to identify types. Adjust callers.
CLanguageFamilySyntaxHighlightInfo:
- Use TeamTypeInformation to check if identifiers map to types, and highlight
accordingly if they do.
TeamTypeInformation:
- Add interface function to allow simply querying for whether a
type is known to exist by a name + constraints, without actually
requesting that it be instantiated, and implement in TeamDebugInfo.
GlobalTypeLookup:
- Add analogous type lookup function to the above, and implement in
TeamDebugInfo.
{Specific,Debugger,Dwarf}ImageDebugInfo:
- Add function for type existence query, and implement in subclasses.
DwarfImageDebugInfo:
- On initialization, we now walk the list of compilation units and
build a hash table of all of their contained types. The table is then
used by GetType() to quickly find the subset of DIEType objects that
match the requested name, and then compare only those to the
given constraints to find the best match.
For a more complex image such as libbe or Debugger itself, this reduces
the time for an uncached type lookup from around 50 msec to < 10 usec on
my i7.
* BGLView has it's own locking, this isn't the correct
behaviour as you only lock for gl* calls
* Resolves problems with window not scaling properly
on proper llvmpipe
- The Backgrounds preferences app will now accept color drops.
- The Backgrounds app will update if a color drop to the desktop
has changed its color due to a direct color drop
Signed-off-by: Jerome Duval <jerome.duval@gmail.com>
* When we get bogus values back as the current battery state, there
is little reason to present this to the user when we can detect it.
* While it would be nice to solve the underlying bug, too (if there is
one, this should make the new low battery notification feature a lot
nicer in practice.
- When resolving an identifier, first attempt to match it against
frame variable if applicable, and only attempt to look it up as a
type name if the former comes up empty. The type lookup is
significantly more expensive since if it doesn't match an already
known type, it currently needs to search through all DIEs trying
to find a candidate. This would significantly slow down expression
evaluation involving identifier names, and was particularly noticeable
for breakpoint condition processing.
- When evaluating a breakpoint condition, there's no need to change
the thread state unless the condition is actually met. This would
lead to lots of unnecessary state switching and associated overhead
in the GUI.
- When detecting a possible start of string literal, ensure that a
terminator was actually found, otherwise simply consider it as a
single character token. Otherwise, cases such as an apostrophe in
a C-style comment that spanned a single line would result in us
missing the token for the comment close, and highlighting all
subsequent lines until such a token was encountered as comments.
Let Model::PopulatePackage() have a mode where it forces the update on a
package that has already been populated. Trigger populating from success
code path in RatePackageWindow. Fixes#11638.
SourceView:
- When calculating line widths for scrollbar calculation purposes, tabstop
offsets weren't being correctly taken into account, which would result
in lines being computed as shorter than they actually were. Depending
on the content of the file, the horizontal scrollbar would consequently
be configured improperly in some cases.
