Generalize the fix of d6fb86a81c and
explicitly activate the parent when a TLW with wxFRAME_FLOAT_ON_PARENT
style is hidden or minimized and not only when it is destroyed.
Closes#18535.
Ensure that the dialog is still alive when it gets the kill focus event
for its child which had focus just before the dialog was closed (or any
other events generated by this child when it detects that it's losing
focus, such as wxEVT_SPINCTRL) by resetting focus when the dialog is
being hidden and not when it's being destroyed.
This makes the events order more consistent with wxMSW but also, most
importantly, safer, as wxEVT_KILL_FOCUS handlers could previously easily
reference the fields of an already half-destroyed wxDialog-derived
object by the time they were run during wxTopLevelWindowGTK destructor
execution.
Closes#18145.
This is yet another fix related to recent wxBMPHandler changes: this one
is trivial, as it just avoids gcc signed/unsigned comparison warning,
but is still needed for the clean build.
See #18634.
The code from the previous commit used sizeof(BITMAPINFOHEADER), but
this struct is only defined under MSW, so this broke the build under the
other platforms.
Luckily, we don't actually need the struct itself, but just its size, so
simply hardcode it here as it's fixed (part of the BMP format) and not
going to change.
See #18634.
Only BITMAPINFOHEADER data are explicitly read and processed so for a bitmap
with BITMAPV4HEADER or BITMAPV5HEADER we have to forward stream position
to the end of the actual header in order to start reading colour table
from the right position.
Closes#18634.
The documentation's notes about MSW limitations about setting priority
before creating the thread do not appear to be true (anymore). Thread
priority is already set by Create() if SetPriority() was called earlier.
Setting it immediately just failed, because the thread did not exist
yet, but this was fixed by the previous commit.
This adds a demonstration of drawing with affine 2D transformation applied
to wxDC. Transformation paramaters (translation, scale, rotation angle)
can be set/reset with new menu "Transformation".
Optimized function to draw H/V lines is based on ExtTextOut() API and therefore raster operations modes other than wxCOPY are not applicable.
Closes#18635.
This script used to live in oss-fuzz repository itself (under
projects/wxwidgets), but it seems better to have it in the main
repository, as this will make modifying it, e.g. to add new fuzzers,
simpler.
Don't trigger a confusing assertion failure about unknown scale mode if
a 0-sized window is being repainted.
It's not clear why should an empty window get repaint events in the
first place, but this does happen, at least under MSW, so guard against
it.
Icons were not supported any longer since the switch to the native
implementation in 235e61c311.
Restore them now by using text attachments (thanks to Scott B for the
idea). The vertical alignment is still not right, but it's better than
nothing.
See #17473.
Instead of using separate check and icon+text columns in the list model
page, use the combined check+icon+text column to verify that it works
(which is not the case current in the native Mac version).
Don't call -[NSCell setTextColor:] unless it really handles it: testing
for "colText" validity was insufficient, as it could be set as a
fallback.
Also move the fallback assignments inside the branches where they're
really used.
This fixes a bug introduced in 540fed9216,
see https://github.com/wxWidgets/wxWidgets/pull/1673
Restrict GTK workaround in wxTreeListModel::IsContainer() to the native
version, it does have a drawback when using the generic one as it
resulted in all items having the expander button.
See #18632.
It defaults to 1, but can be set to 0 to force using the generic version
of the control even under GTK or Mac, where the native version is used
by default.
This can, unfortunately, be useful to work around various but multiple
native control limitations.
The changes of the previous commit were insufficient, as we could also
use an invalid column index when handling mouse event in an empty grid.
Fix this by checking for the line validity in PosToEdgeOfLine() instead
of adding another ad hoc check to wxGridColLabelWindow itself, as it
seems reasonable for this function to verify that it uses valid indices.
See #18629.
When the grid is empty, XToCol() has no choice but to return invalid
index (because there are no valid ones) when it's called from the
drawing code and passing such indices to GetColPos() results in an
assert, which is fatal inside wxEVT_PAINT handler.
Just short-circuit all this code by not doing anything in the case of
empty grid. This is sloppy, but simpler and more robust than any
alternatives.
Closes#18629.
It is said in the wxDC::DrawText() documentation that current logical function is ignored by this function (e928566f). This should apply also to wxGCDC implementation for the sake of consistency and text drawing should be always done with wxCOPY mode regardless of the current mode set.
This makes more sense for a page of a GtkNotebook, which is always
positioned at (0, 0) relative to its client area, and is also more
compatible with wxMSW.
Closes#18548.
Closes https://github.com/wxWidgets/wxWidgets/pull/1670
