* Select items in the encoding list by block identifier rather than
list index (there are holes in the list)
* Scroll to the start of the list when the application starts
* Add and use IsBlockVisible to detect when we need to make a block
visible, avoiding unwanted scrollings in some cases with the previous
method of testing for only one character in the block.
Signed-off-by: Adrien Destugues <pulkomandy@gmail.com>
Fixes some problems with previous enhancements from #3651.
* only update if font family update is detected.
* character map can now detect new fonts or removed fonts without
restarting
Signed-off-by: Adrien Destugues <pulkomandy@gmail.com>
Fixes#3651 part 7.
The computation of the graph width could overflow leading to an absurdly
high count of lines to be drawn. This would freeze with BShape trying to
realloc the array to bigger and bigger sizes as lines were added to it
(by increment of 256, but still), or trigerred a bad_alloc in
BeginLineArray which allocates them upfront after I converted the code
to use that (it is more appropriate).
Add sanity checks to avoid the overflow, now you can have hundreds of
graphs without problems (much more than I can fit on my screen anyway,
which I think is good enough).
Fixes#6841.
We allow this method to throw bad_alloc exceptions as there is no other
way to report errors. However we left the object in a broken state
(maxCount set, but array not initialized) which would crash when calling
either AddLine or EndLineArray.
Initialize the count to 0 before throwing the exception so now
EndLineArray can be called and operations resumed safely after an
allocation failure.
* The key down message is not delivered when completing a dead key, so
act on key up instead.
* Pressing a character after a dead key would leave the keymap in "dead
key" mode, and that mode would be exited only by pressing a dead key
again.
Fixes#5734.
We add a 8 pixel space between the label and the menu, but we shouldn't
do that when the label is empty. This matches BMenuField behavior.
Fixes part of #10734.
set explicit minimum width of character view to be a bit larger than
fCodeView so that there isn't jumping of the pane sizes when fCodeView's
contents are too long
Signed-off-by: Adrien Destugues <pulkomandy@gmail.com>
Fixes#3651 part 9.
* Scrolling the list of characters now selects the currently visible
block in the blocks list.
Signed-off-by: Adrien Destugues <pulkomandy@gmail.com>
Fixes#3651, part 1
The Be Book says this is a hook method, so the default behavior should
be to do nothing (didn't test BeOS, my install is currently broken).
Showing an alert with just the application thread name is less than
helpful anyway, and this was causing crashes.
Fixes#5554.
DiskProbe, DriveSetup, Icon-O-Matic, Installer, Login,
PackageInstaller, Screenshot, Translation Inspector, BootManager,
CharacterMap, CodyCam, BPrintJob
DiskUsage: removed useless translation, used only for layouting purposes
Thanks to KapiX for spotting these problems.
According to the HIG, button labels should be verbs implying actions.
Replace the "default" button with "create public_html", which is what
will happen if you click it.
Thanks to KapiX for the suggestion.
VariablesView:
- Remove unused member variable.
- Add optional presentation name field to ModelNode, which allows
overriding the name of its underlying ValueNode for display purposes.
Used to ensure that a top-level node for an expression always shows
the expression string itself, even when the returned result is a value
node with a different name.
VariablesView:
- ModelNode now acquires a reference to its target variable. This
is necessary because in the case of expression variables, the
model node is ultimately the only one aware of it, and therefore
needs to manage its lifetime.
- AddSyntheticNode() now discriminates between the case where it
needs to create a node child on behalf of the caller versus when
it's given an already created one, as is the case when an expression
resolves to a value node. Consequently it also detects whether
it needs to request value resolution or not, and handles accordingly.
- _AddExpressionNode() now supports value node results and handles
them accordingly.
- Simplify handling of expression nodes. For primitive results, we now
construct a Variable object that represents the expression result, and
then add that as we would any other local variable. This simplifies handling,
and also allows saving/restoration of their view state to be handled the
same as other nodes. Complex expression results aren't yet handled properly,
pending some further work in progress on the evaluator.
ExpressionInfo:
- No longer stores an explicit result type (this is inferred from
evaluation of the expression itself now).
- Introduce class ExpressionResult for returning the result of an
expression computation. This can currently take the form of either
a primitive value, or a value node object.
- Adjust UserInterfaceListener and ExpressionInfo::Listener to take
the above changes into account, and correspondingly adjust all
callers/listeners.
CLanguageExpressionEvaluator:
- Introduce child class Operand. This subsumes the functionality that
was previously in the separate Number class, and can represent a
primitive value, a value node or a type. Also has functionality to
implicity handle type promotion/inferring when performing calculations
between operands.
- Adjust expression parser to operate in terms of Operands rather than
Numbers. This allows a number of improvements, most notably that an
expression can now return a value node as a result rather than only
a primitive number. This capability isn't yet fully used, but paves
the way for future uses such as an expression that evaluates to a data
member, a global variable, or an arbitrary pointer of a particular type.
- Various cleanups/simplifications that were possible as a result of the above
changes.
ExpressionEvaluationWindow/ExpressionPromptWindow:
- Remove type menu field, since the expression API no longer uses it.
Adding/removing expressions in the VariablesView is temporarily disabled,
pending some further rework there to properly handle the new result object.
We were using the same "no space available" message used when the first
partition starts too early. Be more specific here to make it clearer
what the problem is.
Fixes#7087.
Patch by Karvjorm. Picked only the code changes as it seems the Jamfile
changes were not needed or already done.
* Use BObjectList instead of BList
* Add some more directory constants to FindDirectory tests.
