possible bug in extract_h() function definition in C6000_v7.4.1/include/gsm.h

The C6000 definition is OK for C6000 because "unsigned" is 32 bits, whereas it's 16 bits for C5500.  For this reason, C5500 must shift first.

Hi,

I see it produce the wrong results in my application, however. I noticed the errors when recently converting a working C55X module to C6X. The C6X version will cast a signed 32-bit integer to unsigned and then right shift; so, the sign is lost. To my understanding, the intent is to "extract" the signed upper 16-bit value.

Eric

I think "unsigned" should be removed in both. The reason the C5X version works is that the shift is done first. The unsigned cast, being done last, is overruled by the implicit casting to the data type that is to the left of the equals sign in the C code line where extract_h() is actually used. At least that is my conclusion.

Eric

For C5500, that would not help if using the extract_h macro in a context other than assignment to a 16-bit value, such as if(extract_h(x) < 1), which wouldn't work properly for negative x.