This will allow disabling some tests which fail when running only under
Xvfb.
The new function doesn't work automatically because there doesn't seem
to be any way to distinguish it from the usual server, so it just checks
whether wxUSE_XVFB=1 is set in the environment, as is done by Travis CI
build script.
We need to wait until wxDataViewCtrl is realized for its columns widths
to be actually set, as calling gtk_tree_view_column_set_fixed_width()
just queues a resize that will take effect later, as documented in the
previous commit.
This not only cuts down on its size, but will make it simpler to skip
this test in the environments where web access is unavailable in the
upcoming commit.
Verify that pressing Enter in a dialog activates its default button when
a text-like (i.e. wxTextCtrl or wxComboBox) has focus either if it
doesn't have wxTE_PROCESS_ENTER style or if it does, but its handler
skips the event, but not if the style is used and the event is handled.
This ctor just created unusable wxTimerEvent objects (all of the methods
specific to this class would just crash if called on them) and doesn't
seem to be useful at all.
It was added in e47859daeb apparently only
in order to allow using wxIMPLEMENT_DYNAMIC_CLASS() instead of the
previously used wxIMPLEMENT_ABSTRACT_CLASS() for wxTimerEvent, but there
doesn't seem to be any reason to prefer macro over another, and there
are good reasons to not allow creating objects in an invalid state.
The only place where we relied on having default ctor for this event was
in wxEvent::Clone() unit test, so update it to handle wxTimerEvent
specially now that this ctor doesn't exist any longer.
When processing mnemonics, the resulting string should still contain
them and they need to be stripped before measuring its width, but the
code didn't do it.
This didn't prevent the tests from passing, but only due to another bug
in ellipsization code itself, which lost the mnemonics completely. As
this bug is about to be fixed, the test needs to take mnemonics into
account properly now.
This resulted in wrong letter being underlined in wxGenericStaticText
when the mnemonic occurred after "&&" (i.e. an actual ampersand) in the
label.
Add unit test which passes now, but would fail before on the last check.
Also remove the macros used in the test to perform the same tests for
wxStaticText and wxCheckBox and use a helper function instead, making
the code more clear and extensible.
No real changes.
This shouldn't be done unconditionally, trace mask can always be enabled
using WXTRACE environment variable from outside the test.
Also remove a useless tracing message.
The solution with specializing std::iter_swap() for wxString::iterator
was not conforming as the iterator was still not swappable, as it is
required to be.
Fix this by providing std::swap() overload for wxString::iterator, which
is correct and even simpler.
This allows std::reverse(s.begin(), s.end()) work with clang too and
incidentally avoids warnings about the code relying on non-conforming
extensions with MSVS 2017 which were due to the fact that iter_swap()
workaround wasn't enabled for it, while the new swap() overload is.
These functions were never used anywhere and their meaning was not
clear, so just drop them.
Also remove the now unused "tested" member and "WAIT_DURATION" constant.
Use CallAfter() to perform idle-time initialization, this is shorter and
simpler than dealing with the idle events explicitly.
Unfortunately the remaining use of idle events can't be avoided, explain
the reasons for this in a comment.
Just bind the idle events we want to execute dynamically instead.
This is already more clear and robust than the old version, but will be
simplified event further in the next commit.
This reverts commit a05ae051d8 and allows
to correctly decode UTF-8 output of child processes again.
Also add a (disabled by default) test allowing to verify that this does
work now.
Closes#14720, #18382.
Don't assert in wxBoxSizer when both alignment flags and wxEXPAND are
used together if wxSHAPED is also used, as such flag combinations may
make sense and so shouldn't be forbidden.
Add a unit test checking that this is allowed.
Resetting the insertion point position to 0 after calling
wxTextCtrl::SetValue() or ChangeValue() which didn't really change the
control contents was unexpected, as such calls are supposed to be
"optimized away", and this was indeed the case under wxMSW and wxOSX,
but not in wxGTK.
So change wxGTK to follow the other ports, add a unit test checking for
this behaviour and officially document it.
As a side effect, this ensures that the numeric validator classes don't
reset the insertion point position to 0 on every focus loss under wxGTK,
as happened before.
