The original code that this was copied from only disabled this check in
debug builds as described in the comment, however doing so also allows
`channelIndex` to increase without bounds and can lead to writing past
the end of the output buffer.
https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=27503
- Add MPEG major and minor formats to string printing routines.
- Don't warn when adding a broadcast chunk to a MPEG WAV.
- sndfile-convert: Infer that ".ogg" suffix means Ogg+Vorbis. This
aligns with Xiph.org
- sndfile-info: Properly calculate the dB of Vorbis and Opus files. The
handling of the allowable sample range of these files may need to be
investigated in future.
libmpg123 supports decoding MPEG Layers I, II, and III, but we only have
encode support for Layer III. Skip checks for I and II as they don't
have write support.
This reverts commit 2269af8417.
The code which defined nullptr mysteriously broke the Windows build for
SuperCollider. After some investigation I discovered this commit to be the
cause, as MSVC by default does not set __cplusplus to anything greater
than 199711L unless explicitly asked to.
In general, #defining a keyword like nullptr in a library header is not
courteous, as it can affect compilation of later code and cause
significant confusion. This was exactly the situation I encountered, and
it took several days to track down the source.
The stated reason for the commit reverted by this was to prevent a
specific warning from appearing during compilation. However, there are
a few less intrusive ways of accomplishing this. For example, by using
`#pragma`s offered by all major compilers which temporarily disable
warnings, or by including headers with `-isystem` or any similar
mechanism which disables warnings from "system" headers. For code which
is intended to support C++98 and newer standards, preventing all
compilers from emitting any warnings is a losing battle. It is easier
in my view for consuming code to handle this as it needs to.
There are several options here for resolving this. I could just fix my
project, but that would not prevent others from encountering the same
issue. I could try for a clever-er version of this using a macro not
named nullptr and an additional check for MSVC, but this seems to offer
diminishing returns considering how easily any warnings about NULL can
be silenced, and carries the risk of introducing new and unforeseen
problems. So, I have here simply reverted the commit so the code uses
NULL once again.
