This avoids stale data presented to caller via a fread()
Current bug due to the buffered data can be seen in stripping mp3s of tags
f.strip(ID3v1);
f.strip(APE);
The ID3v1 tag sits at the end of file (strip calls ftruncate()) and the APE
strip performs a readFile() that would return the stream buffered/truncated data
and reinsert
* Respect atom type when converting rate tag
TagLib prior to #818 (commit ff28cf276c) read and wrote the "rate" tag as
text, and switched to reading it as integer value even when the atom class
is a text type. This breaks reading existing files, and can be avoided by
taking the atom type into account.
This fixes issue #885.
* Respect MP4::Item type when writing the rate tag
TagLib prior to #818 (commit ff28cf276c) read and wrote the "rate" tag
as text, and switched to writing the integer value of the MP4::Item.
This breaks writing from applications which supply the value as
StringList, which was the previously implemented API. Applications using
an MP4::Item(UInt) are still supported without changes on the application
side.
This is the complementary writing part for issue #885.
When parsing mp4 media header version 1 (mdhd) atoms, the timescale (unit) is parsed as a `LongLong` (8 bytes), but instead should be a `UInt` (4 bytes). This results in an incorrect timescale, and also pushes the offset of the duration (length) off by 4 bytes.
The end result being that the AudioProperties track length for mp4's with mdhd v1 comes back as 0.
See: https://wiki.multimedia.cx/index.php/QuickTime_container
| Entry | Bytes (v0) | Bytes (v1) |
| :--- | :---: | :---: |
| size | 4 | 4 |
| type | 4 | 4 |
| version | 1 | 1 |
| flags | 3 | 3 |
| creation time* | 4 | **8** |
| modification time* | 4 | **8** |
| time scale | 4 | 4 |
| duration* | 4 | **8** |
| language | 2 | 2 |
| quality | 2 | 2 |
* Add DSF and DSDIFF file types management
* Fixes for some build chains
* unit64_t replaced by unsigned long long, warning fixes
* Remove C++11 extension incompatible with some build chains (enumeration in a nested name specifier)
* Change typedef types (uint, ulong, ...) to standard types
remove BUILD_FRAMEWORK changes from this pull request
* Replace deprecated String::null and ByteVector::null by String() and ByteVector()
Styling update, thanks to FestusHagen
* Restyling
* Restyling to reduce length of excessively long lines
* Add to detectByExtension
* Added `isSupported(IOStream *stream)` to `DSF::File` and `DSDIFF::File`
CVE-2018-11439 is caused by a failure to check the minimum length
of a ogg flac header. This header is detailed in full at:
https://xiph.org/flac/ogg_mapping.html. Added more strict checking
for entire header.
If TDRC is encrypted, FrameFactory::createFrame() returns UnknownFrame
which causes problems in rebuildAggregateFrames() when it is assumed
that TDRC is a TextIdentificationFrame
powerpc is a platform with 'char' == 'unsigned char'.
As a result '-1' is not expressible in char and build fails as:
```
# '-funsigned-char' to force test build failure on other platforms
$ cmake .. -DBUILD_TESTS=YES -DCMAKE_CXX_FLAGS="-O2 -funsigned-char" -DCMAKE_C_FLAGS="-O2 -funsigned-char"
...
$ make check
tests/test_synchdata.cpp: In member function 'void TestID3v2SynchData::testToUIntBroken()':
tests/test_synchdata.cpp:78:33: error: narrowing conversion of '-1' from 'int' to 'char' inside { } [-Wnarrowing]
char data[] = { 0, 0, 0, -1 };
^
```
The fix is to expliticly cast -1 to 'char'.
Signed-off-by: Sergei Trofimovich <slyfox@gentoo.org>
Don't rely on _WIN32_WINNT value to enable WinRT support.
if _WIN32_WINNT is not set manually, it is defaulted to SDK version. So
if you use SDK > 8 you cannot use TagLib under Win7 and lower because of
CreateFile2 function dependency.
PLATFORM_WINRT option (OFF by default) was introduced to enable WinRT
build.
Related issues: https://github.com/Microsoft/vcpkg/issues/1240
The name 'isValidStream' is a little misleading because it doesn't check if the stream is really valid. Additionally, 'isSupported' can be naturally overloaded.
