Information about display cached in wxDisplayFactory::m_impls and in
wxDisplayFactoryMSW::m_displays could get out of sync after removing a
display from the system, resulting in failures when calling various
wxDisplay functions later.
To fix this, reuse InvalidateCache() to invalidate the cache in
wxDisplayFactoryMSW too, making it virtual in order to allow overriding
it there. Also call InvalidateCache() from wxDisplayFactoryMSW itself
instead of doing it from wxWindow, as this works even when the
application isn't showing any windows (and also keeps all
display-related code together).
Closes https://github.com/wxWidgets/wxWidgets/pull/1246
No real changes, just clean up sources by removing trailing spaces from
all the non-generated files.
This should hopefully avoid future commits mixing significant changes
with insignificant whitespace ones.
The cache added in 990c8bfd73 was not
invalidated properly, meaning that wrong information was returned when
displays were [dis]connected after the application startup.
Fix this at least for MSW by invalidating the cache on receiving
WM_DISPLAYCHANGE (which means that sometimes we will do it
unnecessarily, as the change in resolution of an existing display
doesn't require cache invalidation, but this shouldn't be a big problem
in practice as the speed with which the user can change the display
resolution is not very high).
Closes https://github.com/wxWidgets/wxWidgets/pull/1090
We need to account for the scale factor under GTK+ (and, presumably,
under macOS) to compute the correct PPI value as it must use the number
of physical and not logical pixels.
While this is not done for all the ports yet, the new API allows
returning different PPI values for different monitors, unlike the old
(and still existing, but implemented in terms of the new one) global
function.
Allow getting the depth of any display, not just the primary one, even
though this is not implemented for Unix ports currently.
Mostly do this for consistency with the other display-related functions.
Simplify the code by not making this function pure virtual as all the
ports except MSW had to override it just to return an empty string.
Instead, just return empty string by default as it's not critical to
force the derived classes to override this function.
Instead of forwarding to these functions from wxDisplay implementation
in wxUSE_DISPLAY==0 case, make the functions themselves wrappers around
wxDisplay, which may, or not, depending on the platform, have a simpler
implementation in wxUSE_DISPLAY==0 case, but is always available in any
case.
As part of this change, only use src/osx/core/display.cpp in macOS
builds, not iOS ones and update the Xcode project accordingly too.
This cuts down on code duplication, especially in wxGTK, and facilitates
further additions to wxDisplay API.
Avoid a heap allocation on every wxDisplay creation by caching the
wxDisplayImpl objects once they're created.
Currently we never invalidate the cache, but we should add a way to do
it in the future.
This speeds up wxDisplay::GetGeometry() benchmark by a factor of 4.
Implement the new wxFont pure virtual methods in this port version of
this class.
Neither arbitrary weights nor fractional point sizes are actually
supported in this port however.
Changes of c0b0562533 to common code broke
wxGTK1 build, as wx/gtk/private/wrapgtk.h is for wxGTK 2+ only.
Fix this by handling wxGTK 1 separately and including gtk/gtk.h directly
for it.
Hopefully this code will be removed, together with the rest of wxGTK1
support, in some not so distant future.
Save both the normal window geometry and its maximized position instead
of saving just its current position. This fixes restoring geometry of
the maximized windows as previously they were always restored on the
primary monitor, as their original position was lost.
Use the native {Get,Set}WindowPlacement() functions for a MSW-specific
wxTLWGeometry implementation to achieve this.
Closes#16335.
Previously, TLW geometry was implicitly defined as just its position,
size and the maximized/iconized state by wxPersistentTLW code. This
already wasn't enough for wxGTK which added the decoration sizes to the
geometry being saved/restored, but this had to be done using conditional
compilation, which was not ideal. And it didn't allow using an entirely
different geometry representation as will be done for wxMSW soon.
Change the code to use wxTLWGeometry class defining the geometry, as
used by the current port, explicitly and move wxPersistentTLW logic into
it, as wxPersistentXXX classes are supposed to be very simple, which
wasn't really the case.
Also provide public SaveGeometry() and RestoreToGeometry() methods in
wxTopLevelWindow, which can be useful even to people not using
wxPersistentTLW for whatever reason.
There should be no changes in behaviour so far.
Escape backslashes in wxJSScriptWrapper to allow strings with
backslashes in them to work again -- this was broken while implementing
support for returning values from JavaScript to C++.
See https://github.com/wxWidgets/wxWidgets/pull/741
Due to the same problem with sizeof(long) mismatch as in the previous
commit, we can't use struct timeval, with its long fields, in 64 bit
Cygwin builds, and need to use __ms_timeval instead.
Add wxTimeVal_t type to hide this difference and update all code
compiled under MSW (there is no need to uglify Unix-only code using
timeval, as in wxSelectDispatcher, for example) to use it instead of
timeval.
Allow wxPrintf("%1$s %1$s", "foo") to work.
Take into account the possibility that the number of format specifiers
and the number of actual arguments can be different.
Closes#9367.
No real changes, but just use string concatenation instead of line
continuation backslashes and also replace hard TABs with spaces.
Notice that the code is still completely unreadable as JavaScript, but
this shouldn't be a problem as nobody will see it there anyhow, so it's
quite enough to have it more readable in its string form in C++.
We can prepare the escaped code directly here, instead of waiting for
GetWrappedCode() call, this is more clear and safer as it avoids
escaping the code twice accidentally if GetWrappedCode() ends up being
called twice somehow.
While it is allowed there, it is confusing and unnecessary.
Also improve the comment explaining why do we need this output variable
name at all and why does it need to be unique.
This is a non-standard extension in the first place and the variable
name didn't make any sense too, so rename it to be standard-conforming
and actually correspond to what it contains.
The old API didn't make any sense for the most common case when both the
user name and password need to be stored, as it required providing the
user name as input, which couldn't work (but somehow this went
unnoticed for more than a year...).
Fix this by returning the username, and not only the password, from
Load() instead of taking it as parameter and removing this parameter
from Delete() as well.
Also improve the documentation, notably include a simple example of
using this class.
Notice that this is a backwards-incompatible change, but the old API was
really badly broken and didn't appear in 3.1.0 yet, so the breakage is
both unavoidable and, hopefully, shouldn't affect much code.
Nevertheless, a special wxHAS_SECRETSTORE_LOAD_USERNAME symbol is added
to allow testing for it if necessary.
Fix the logic for restoring previous span's colors to account for the
possibility of spans that don't change the color, such as in "<i><span
color='red'>...</span></i>foo". Previously, "foo" would always be
rendered black, because unwinding the attributes stack would encounter
an invalid color (which has r=g=b=0) and set it, disregarding and
preexisting attributes.
Because there's code in there that checks whether the attributes are
valid, we need to keep track of both the actually specified attributes
and the currently effective ones, and use the latter for restoration.
Add a new class allowing to store passwords and other sensitive information
using the OS-provided facilities.
Add implementations for all the main platforms, documentation and a new sample
(which contains an ad hoc unit test as the real unit test for this class would
probably be a bad idea as it wouldn't run in non-interactive contexts and
could show OS level dialog boxes if it did).
Extract platform-specific code in a wxUIActionSimulatorImpl-derived class
instead of keeping it in wxUIActionSimulator itself.
This will allow determining which implementation to use dynamically (i.e. at
run-time and not compile-time) to use later and already allows to get rid of
an __WXOSX__ #ifdef in common code.
We reuse the same global dispatcher object (allocated in
wxFDIODispatcher::Get()) for the sockets created in different threads, so it's
perfectly possible for its methods to be called concurrently and this happens
even in our own socket streams unit tests.
Protect against concurrent modification of the select sets and m_maxFD. This
fixes sporadic Travis build failures such as the one at
https://travis-ci.org/wxWidgets/wxWidgets/jobs/110757281 for example and
probably even worse bugs too.
wxNotificationMessage has been refactored to always use wxNotificationMessageImpl (this was previously already done in the MSW implementation)
This adds various features and fixes to wxNotificationMessage:
- OS X Notification Center implementation
- Generic "toast" notifications
- SetIcon() to specify a custom icon
- AddAction() to add actions to notifications
- Events to get notify of notification clicks, dismiss or actions
