Don't define operators for wx types in global scope.
This results in much better error messages when the compiler gives an
error about failing to resolve a call to some operator in user code
because it doesn't list all the wx-defined operators as candidates. In
practice, this cuts down error messages from 500 (!) lines to 2 with
gcc.
See #24190.
This was probably done unintentionally in 8ccbd7e95d (Implement
operator/=(int) and operator*=(int) for wxPoint and wxRealPoint.,
2024-01-04) and results in warnings from clang.
This function should be used instead of CreateWithDIPSize() in portable
code, the latter is currently often used to do what this function is
doing, but this only works correctly in wxOSX and wxGTK3 but not wxMSW.
Still use the traditional dot file if it already exists, but prefer
using the new location otherwise, i.e. for the new program installation.
Add wxCONFIG_USE_HOME to allow forcing the use of the old location if
really necessary.
Also use the new style as default "old style" of MigrateLocalFile() so
that calling it even when using XDG layout in wxStandardPaths still
works as expected.
As not defining operator<<() overload taking wxScopedCharBuffer in the
global scope prevents it from being considered as an overload resolution
candidate (which is, of course, the whole point), it also prevents it
from being used for the classes convertible to it, such as wxCharBuffer,
so we need to define operator<<() overloaded for the latter explicitly
too.
We also need a new wxScopedCharTypeBufferStreamSupport helper in order
to define different operators inside different specializations of
wxScopedCharTypeBuffer<>.
Add new wxDECLARE_VARIANT_OBJECT_EXPORTED() macro defining these
operators as friend functions inside the class declaration and replace
all uses of DECLARE_VARIANT_OBJECT_EXPORTED() inside the core library
with the new macro to avoid defining any operator<<() overloads in the
global scope.
Also add wxIMPLEMENT_VARIANT_OBJECT() for consistency, even though it is
not really needed.
Use "hidden friend" idiom instead and define these operators as friend
functions inside the corresponding class scope, so that they're only
found using ADL and, in particular, don't appear as candidates when
looking for any operator.
In practice, this significantly reduces the error messages given if some
operator (e.g. "==") is applied to a type not defined it, as the
compiler doesn't need to consider converting this type to wxPoint,
wxRealPoint, wxRect, wxSize etc, nor to complain about failing to do it.
The arithmetic operators for wxPoint and wxRealPoint are quite
repetitive and were made by copy-paste from the wxSize operators.
Because of that, some of the names are a bit misleading which this
commit changes. The changes are:
1. replace s/sz with p/pt for point variables (likely 's' comes from
copied code used for wxSize variables)
2. replace 'i' with 'f' for floating point types (likely 'i' comes
from copied code used for integer types)
3. replace 'factor' with 'divisor' for division operations (factors
are the multiplication operands, not division)
We need to do it when using XWayland for the same reasons as we had to
do it in the EGL version when using either XWayland or Wayland directly:
without this, we can block for up to 1 second in glXSwapBuffers() if the
window is hidden, see #23512.
Closes#24163.
Closes#24165.
Do this for consistency with wxSize which already had this operator/()
overload and operator*() in these classes which was already overloaded
for double.
Closes#24185.
Closes#24187.
This can be useful for the existing applications switching to using
XDG-compliant config files location as they can just call this function
on startup to keep using the existing file.
Sometimes it's useful to have some code used only if wxConfig is defined
as wxRegConfig or only if it is defined as wxFileConfig and testing for
these symbols is more clear than testing for the platform (and more
correct, considering that setting wxUSE_CONFIG_NATIVE to 0 may result in
wxFileConfig being used even under Windows).
No real changes, just add a function ensuring that the tab is shown.
It's not really clear why the existing code sometimes calls
MakeTabVisible() after SetActivePage() and sometimes doesn't, and
perhaps it should actually always do it, in which case this function
could be merged with SetActivePage() to do everything at once, but for
now keep the changes minimal.
Using this style allows to tell wxFileConfig to use XDG-compliant
location for the user configuration file, even when not using XDG file
layout for wxStandardPaths, which can be convenient when modifying the
existing application using wxStandardPaths for other files locations
too.
And using it in combination with wxCONFIG_USE_SUBDIR allows to put the
config file in a XDG-compliant subdirectory, which wasn't easily
possible at all before.
It doesn't make sense to call this function and ignore its return value,
but it may happen accidentally as it might seem that it modifies its
parameter.
This allows using GetUserDir(Dir_Config) to retrieve XDG-compliant
location of the configuration files, independently of the file layout
used by the global wxStandardPaths object.
Recommend using this directory instead of GetUserDataDir() for multiple
configuration files under Unix.
This parameter wasn't used for anything as it specified the encoding of
the data contents in memory and not the encoding of the XML file as
might be believed, so it only created unnecessary confusion.
See #24167.
