Overflow in channel convertion

[tanguimorvan]

Actual behavior

Channel conversion from boost::gil::packed_channel_value to uint16_t fails due to integer overflow in channel_converter_unsigned_integral_nondivisible<SrcChannelV,DstChannelV,true,false> .

Expected behavior

The value should not overflow, producing correct results.

C++ Minimal Working Example

#include <boost/gil.hpp>
namespace gil = boost::gil;
int main
{
    boost::gil::packed_channel_value<10> packedChannelValue(3);
    //Expected 192, got 64
    uint16_t convertedValue = boost::gil::channel_convert<uint16_t>(packedChannelValue);
}

Environment

• Compiler version: GCC 8.4.0
• Build settings:
• Version (Git ref or <boost/version.hpp>): 1.68 but seems to still be present in latest version.

[tanguimorvan]

Index: channel_algorithm.hpp
===================================================================
--- channel_algorithm.hpp	(revision 56209)
+++ channel_algorithm.hpp	(working copy)
@@ -237,7 +237,7 @@
 struct channel_converter_unsigned_integral_nondivisible<SrcChannelV,DstChannelV,true,false> {
     DstChannelV operator()(SrcChannelV src) const {
         typedef typename base_channel_type<DstChannelV>::type dest_t;
-        return DstChannelV(static_cast<dest_t>( src * unsigned_integral_max_value<DstChannelV>::value) / unsigned_integral_max_value<SrcChannelV>::value);
+        return DstChannelV(static_cast<dest_t>( src ) * unsigned_integral_max_value<DstChannelV>::value / unsigned_integral_max_value<SrcChannelV>::value);
     }
 };

[mloskot]

@tanguimorvan Thanks, feel free to open PR with your changes