Subtracting a huge negative offset from the current position could
overflow it, which was correctly detected as undefined behaviour by
UBSAN.
Credit to OSS-Fuzz: this solves its issue 4388.
Although the object being created _will_ be of type wxAppConsole, it
doesn't have this type yet, as wxAppConsole ctor hadn't yet run, and so
the static cast is invalid and detected as such by UBSAN in g++ 6.3.
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.
Probably a follow-up fix for commit from 10 years ago: 288b610
This fixes compilation with `--enable-extended_rtti` configure flag.
Also removed all trailing spaces in this file.
Replaced `wxToStringConverter<type, data, result)` with
`wxFromStringConverter<type>(data, result)`, which was broken
for 7 years since the commit 13830d6
This fixes compilation with `--enable-extended_rtti` configure flag.
Also removed all trailing spaces in this file.
Handle this feature as all the other ones and provide a configure switch
and a setup.h option to disable it if necessary, as it may be desirable
to do it, especially under Linux, to avoid extra dependency on pangoft2
if this functionality is unnecessary.
Remove maximal reallocation size in wxArrayString too, as it was done
for wxVector a commit ago, and increase its size by 100% and not 50%
when it needs to grow.
There is no real reason to use different growth strategies in the two
classes and wxVector one seems to be better.
Call assign() instead of Add() in a loop: this is not only shorter, but
also ensures that reserve() is called before starting the loop and all
the required memory is allocated at once.
Just "activate" the font immediately when adding it using
AddPrivateFont(), nothing seems to be gained from having two functions
and it just makes things more complicated both when implementing and
when using the API.
This is necessary to recognize new fonts, for example those added using
AddPrivateFont(), as being valid: otherwise they were not found in the
cache and couldn't be used in wxGTK.
Don't rely on it being done during statics cleanup as this doesn't work
if the library is shutdown and re-initialized.
Use a module to do the cleanup, just as it's already done for a lot of
other global data in wx.
wxFont::AddPrivateFont() can now be used to load a font from a file for the
applications private use. Update the font sample to show this.
Closes#13568.
There don't seem to be any really important changes affecting our use of
the library, but it seems better to use the 2016 version instead of the
1998 one and it also allows to get rid of many hacks that were needed
before as the things they worked around are now taken care of by the
library itself.
Contrary to what a comment in wxTextInputStream::GetChar() said, it is
actually possible to get more than one wide character from a call to
wxMBConv::ToWChar(len+1) even if a previous call to ToWChar(len) failed
to decode anything at all. This happens with wxConvAuto because it keeps
returning an error while it doesn't have enough data to determine if the
input contains a BOM or not, but then returns all the characters
examined so far at once if it turns out that there was no BOM, after
all.
The simplest case in which this created problems was just input starting
with a NUL byte as it as this could be a start of UTF-32BE BOM.
The fix consists in keeping all the bytes read but not yet decoded in
the m_lastBytes buffer and retrying to decode them during the next
GetChar() call. This implies keeping track of how much valid data is
there in m_lastBytes exactly, as we can't discard the already decoded
data immediately, but need to keep it in the buffer too, in order to
allow implementing UngetLast(). Incidentally, UngetLast() was totally
broken for UTF-16/32 input (containing NUL bytes in the middle of the
characters) before and this change fixes this as a side effect.
Also add test cases for previously failing inputs.
No real changes, but just get rid of two functions doing the same thing
but using (semantically) different API, this was just too confusing.
Change all the code to use wxDecodeSurrogate() that encapsulates
decoding the surrogate and advancing the input pointer as needed and so
is less error-prone.
More generally, change the code to use end pointers instead of
decrementing the length to check for the end condition: this is more
clear, simpler and probably even more efficient.
Pass length value to decode_utf16() and end pointer to
wxDecodeSurrogate() to ensure that we never read beyond the end of the
buffer when decoding UTF-16 when the last (complete) 16 bit value in the
buffer is the first half of a surrogate.
This had been previously partially addressed by ad hoc changes, e.g.
f72aa7b1c9 did it for wxMBConvUTF16swap,
but the problem still remained for wxMBConvUTF16straight. Ensure that
this bug is fixed everywhere now but making it impossible to even try
decoding a surrogate without providing the buffer length.
Under Unix systems, this is the same thing, but under MSW, where
sizeof(wchar_t) == 2, this allows to pass wchar_t pointers to this
function without casts.
It also makes it consistent with wxDecodeSurrogate() and allows to get
rid of another ugly cast there.
No real changes.
Calling wxMBConvUTF7::ToWChar(..., "+", 1) resulted in reading
uninitialized memory as the decoding code didn't check that there were
any bytes left when switching to the "shifted" mode.
Fix this by explicitly checking for this and returning an error if
nothing is left.
Having NextChar() returning wxEOT only for GetChar() to turn it back to
NUL didn't make any sense, just return NUL directly and get rid of
GetChar/NextChar() distinction.
No real changes, just simplify the code.
Blindly adding 4 to an unsigned length field could result in 0 and this
led to an infinite loop while iterating over all header fields.
Avoid this by promoting the length to int first, before adding 4 to it.
Credit to OSS-Fuzz: this solves its issue 4083.
Add wxMessageOutputWithConv mix-in class to avoid duplicating the same
code in wxLogStream and wxMessageOutputStderr.
Also derive wxLogStderr from wxMessageOutputStderr to reuse its code
without having to create a temporary object of this type (which will be
more expensive now that doing it involves creating a heap-allocated
conversion object copy).
This commit refactors the overloads of cMB2WC() and cWC2MB() methods
taking raw pointers and buffers to reuse the same code and fixes the
wrong length of the buffer returned by cWC2MB(wchar_t*) overload for
conversions using multiple bytes to represent the NUL terminator
character (it previously was wrong for UTF-16 and UTF-32 conversions due
to wrongly subtracting 1 from the length when creating it instead of
correctly subtracting GetMBNulLen()) and the wrong length of the buffer
returned from cMB2WC(char*) overload where no adjustment for the
trailing NUL was done at all.
Also return simple default-constructed buffers from these methods in
case of failure instead of using wxScopedCharBuffer::CreateNonOwned()
which is less obvious and less efficient (even if the latter probably
doesn't matter here because it's only done in case of an error).
Finally, add tests checking that using WC2MB() or either of cWC2MB()
overloads returns the buffers of the same length and with the same
contents.
We can only use utf8_length() if the conversion used with this object
uses UTF-8 too, otherwise we still need to do the conversion to find out
how many bytes does the string content take in the given encoding.
See #17985.
The Darwin linking problem mentioned in the comment doesn't exist in any
of the still supported macOS versions, so it doesn't make sense to
continue working around it.
Initialize m_pos correctly when using an existing, and hence possibly
not empty, string (and not the internal one which is always empty
initially). The old code was totally wrong as it divided the string
length by the size of wxChar instead of multiplying by it, but doing
this could have been wrong too with UTF-16 and surrogates, so use the
conversion object to compute the real length of the string
representation in the corresponding encoding.
Add a simple unit test checking that this works as intended.
Closes#17985.
Because top windows can (and do) act as parents for certain dialogs, a
window already pending delete shouldn't be explicitly reported as the top
window because all dialogs which would use it as a parent
would be destroyed at nearest idle cycle.
Closes#17982.
Check for the record size before subtracting it from the end position:
the former must be smaller than the latter for all valid ZIP files and
not performing this check could result in an integer overflow error from
the undefined behaviour sanitizer for bad input.
Credit to OSS-Fuzz: this solves its issue 3828.
Zip filenames containing non ASCII characters will be marked with
bit 11 in the general purpose flags and will use UTF-8 encoding.
By only setting the flag when non ASCII characters are used the
created archives should be binary identical to previous versions.
The old behavior can be achieved by explicitly using wxConvLocal
with the constructor. This should also ensure that
existing code using a custom wxMBConv should work as before.
Skip memcpy() call if its source and destination would overlap: this is
not allowed and is correctly flagged as an error by address sanitizer
and is unnecessary anyhow as we're certainly not going to find the magic
value in fewer than 3 remaining bytes.
Credit to OSS-Fuzz: this solves its issue 3794.
Provide native implementation of this function instead of using the ad hoc one
in common code, which didn't really work -- so remove it completely now.
Closes#17189.
Fix comparing items with checkboxes in wxTreeListCtrl and make it
simpler to correctly implement item comparison in other
wxDataViewCtrl-derived classes.
See https://github.com/wxWidgets/wxWidgets/pull/558
This allows to hide the long (and possibly not fitting on the screen)
call stack by default to avoid intimidating people not used to it and
provides a much more clear way to ignore the subsequent asserts, by
clicking a dedicated checkbox instead of having to choose the "Cancel"
button which didn't make much sense.
See #15430.
This makes more sense than the default default (sic) "Yes" button,
pressing which accidentally could kill the program if not running under
the debugger.
This should have been part of 55fd62c1e3
which only updated the default assert handler, but not the one used by
default in all GUI applications, for some reason, see there for more
explanations.
Do this now to ensure that after pressing "Yes" in the assert failure
dialog, the debugger opens at the assert location and not deep inside
wxWidgets code.
See #11184.
Export this class, which was only used internally by wxTreeListCtrl
before, so that user code can use it for its own columns with custom
wxDataViewCtrl models.
