[vlc-devel] [PATCH] contrib: sidplay2: fix signed char stored as int8_t

[Martin Storsjö]

On Thu, 9 Jan 2020, Steve Lhomme wrote:

> Fixes compilation with gcc 9 for Raspberry
> ---
> contrib/src/sidplay2/rules.mak                |  1 +
> contrib/src/sidplay2/sidplay2-char-cast.patch | 22 +++++++++++++++++++
> 2 files changed, 23 insertions(+)
> create mode 100644 contrib/src/sidplay2/sidplay2-char-cast.patch

Um, can you quote the actual error you're fixing here? Because intuitively 
to me, the current state seems much more correct.

>
> diff --git a/contrib/src/sidplay2/rules.mak b/contrib/src/sidplay2/rules.mak
> index 84c35ccfee7..de77935774a 100644
> --- a/contrib/src/sidplay2/rules.mak
> +++ b/contrib/src/sidplay2/rules.mak
> @@ -25,6 +25,7 @@ sidplay-libs: sidplay-libs-$(SID_VERSION).tar.gz .sum-sidplay2
> 	$(APPLY) $(SRC)/sidplay2/sidplay2-string.patch
> 	$(APPLY) $(SRC)/sidplay2/sidplay-fix-ln-s.patch
> 	$(APPLY) $(SRC)/sidplay2/sidplay2-resid-dependency.patch
> +	$(APPLY) $(SRC)/sidplay2/sidplay2-char-cast.patch
> 	$(MOVE)
> 
> .sidplay2: sidplay-libs
> diff --git a/contrib/src/sidplay2/sidplay2-char-cast.patch b/contrib/src/sidplay2/sidplay2-char-cast.patch
> new file mode 100644
> index 00000000000..a16898945fa
> --- /dev/null
> +++ b/contrib/src/sidplay2/sidplay2-char-cast.patch
> @@ -0,0 +1,22 @@
> +--- sidplay-libs/libsidplay/src/xsid/xsid.cpp.char	2020-01-08 14:09:08.124096200 +0100
> ++++ sidplay-libs/libsidplay/src/xsid/xsid.cpp	2020-01-08 14:09:16.204111900 +0100
> +@@ -94,7 +94,7 @@ const int8_t XSID::sampleConvertTable[16
> +     '\x08', '\x19', '\x2a', '\x3b', '\x4c', '\x5d', '\x6e', '\x7f'
> + };
> + */
> +-const int8_t XSID::sampleConvertTable[16] =
> ++const char XSID::sampleConvertTable[16] =
> + {
> +     '\x80', '\x94', '\xa9', '\xbc', '\xce', '\xe1', '\xf2', '\x03',
> +     '\x1b', '\x2a', '\x3b', '\x49', '\x58', '\x66', '\x73', '\x7f'

Using 'char' here is probably wrong, as the plain 'char' type has 
different signedness on different platforms.

Is the problem that the table contains values that are intended to be 
interpreted as signed, but written in the table as unsigned values?

// Martin