Somehow more files appear as modified now, after fixing "all" of them in the
previous commit. Convert these files as well. Will there be more in the next
clone?
This fixes compilation in C++11 mode: "override" results in an error on
non-Mac platforms because this method doesn't exist there and so can't be
overridden.
Closes#16903.
Set bNeedsUpdate flag for all columns even when the control is already empty,
this ensures that things will work if more items are added to it later.
See #10326.
Having CRLF files in the repository with eol=crlf breaks just about
everything, e.g. any rebase/merge involving these files would fail.
Try to fix this by making sure the files use LF in the repository and CRLF is
only used in the working copy.
New element can be added to the list not only by clicking "New item" button but also by clicking under the last element of the list and we need to determine its index in both cases because it is used later on in wxPGArrayEditorDialog::OnEndLabelEdit event handler.
Closes#16905.
Use slashes, not backslashes, in the paths.
Use "auto" for all text files under docs/msw as all of them could be opened in
Notepad.
Also use Unix EOLs for configure.in (and not just for configure).
g++ 4.9.2 added support of __has_include() but, unlike clang, refuses to
compile the <type_traits> header that is detected as existing now in C++98
mode, so the build was broken with it when not using configure (i.e. under
MSW).
Fix this by, first, testing for C++11 compilers separately (which seems like a
good idea anyhow as it will allow using these headers with other compilers)
and, second, not trusting g++ __has_include() for C++11 headers in C++98 mode.
Test using wxTextCtrl in the sample too to demonstrate that currently we need
to get the text contents from the event handler, retrieving it once the dialog
closes doesn't work, at least under MSW.
Fix possible deadlock in the child which would happen if it continued to
produce output after the parent, that used wxExecute() with IO redirection to
launch it, exits.
Do this by preventing the child from inheriting the read ends of the standard
output and error pipes, just as it was already done for the write end of the
standard input pipe.
Closes#16898.
Simplify the code in wxExecute() by using SetHandleInformation() to make the
handle non-inheritable instead of duplicating it to achieve the same goal (the
old code worked under Windows 9x too, unlike the new one, but we don't care
about them any more).
See #16898.
Use both the contents and the header width when wxLIST_AUTOSIZE_USEHEADER is
given instead of just the latter.
Also make both wxLIST_AUTOSIZE_USEHEADER and the previously implemented
wxLIST_AUTOSIZE work efficiently for the virtual list controls by reusing
wxMaxWidthCalculatorBase already used in the generic wxDataViewCtrl.
Closes#10326.
When launching child processes it can be convenient to be able to switch to
them later, provide a method in wxProcess to do it.
Currently this is only implemented in wxMSW but could almost certainly be done
for wxOSX too (it can be done using Apple Script, so presumably there is a way
to do it programmatically as well) and could be also made to work at least
under some Unix systems by emulating what wmctrl does (or just launching it?).
Don't suspend the logs forever if showing the log dialog throws an exception
(this is unlikely but may happen when running unattended GUI tests or using
the dialog hooks throwing exceptions for any other purpose).
Use AC_CACHE_CHECK() to avoid recompiling the test program(s) every time
configure runs and also simplify the checks. This required renaming
ac_wxgtkXXX to wx_cv_gtkXXX (cached variables must have "cv" in them).
Also make the checks more logical, by simply setting wx_cv_gtkXXX if
wx_cv_gtkYYY is set with XXX < YYY.
Finally, and the real reason for these changes, add test for __WXGTK220__
which will be used by the upcoming commits to check whether GtkSpinner is
available.
It was confusing to not see anything when selecting one of the screens using
wxGraphicsRenderer if a renderer hadn't been selected from the menu before.
Switch to using wxGC automatically now if one of these screens selected and
there was no previously selected renderer.
Don't use NSCriticalAlertStyle for wxICON_WARNING unconditionally, this
was a violation of the OS X guidelines. According to the hig HIG,
NSCriticalAlertStyle (aka caution icon) is only appropriate in rare
cases and only if the user is performing a task that might result in the
inadvertent and unexpected destruction of data.
It therefore doesn't make sense to use it for information warnings, but
only for _questions_, and so NSCriticalAlertStyle is only used if a
combination of wxICON_WARNING with either wxYES_NO or wxCANCEL is used.
Make wxICON_WARNING and wxICON_ERROR the primary constants for their
purpose and define their visual Windows-based synonyms wxICON_HAND and
wxICON_EXCLAMATION in terms of the semantic ones, instead of the other
way around.
This is useless at best and resulted in GTK+ warnings because we ended up
(somehow -- is there another bug lurking here?) with negative items height in
this case.
Closes#16862.
Round the double value before passing it to wxColour::Set() instead of relying
on the implicit conversion which is dangerous and correctly flagged as such by
g++.
When the TZ offset specified in the RFC 822 string was equal to the local TZ
offset but the date fell in the DST period, the result was one hour off.
Fix this by converting the date to the UTC explicitly, and then converting it
back to the local TZ to ensure that the DST is taken into effect.
See #15370.
This was broken due to a wrong cast from WPARAM (i.e. 32 bit unsigned int in
Win32) value containing the ID to WORD (i.e. 16 bit unsigned short) which
truncated the ID value.
Notice that MSDN documents the requirement for the IDs to be positive, but
negative IDs seem to work, at least with Windows 7, after this fix.
Closes#16880.
