Add past change on negative LPC shift (#191)

flac.md

@@ -659,7 +659,7 @@ Data             | Description
 `s(n)`           | Unencoded predictor coefficients (n = predictor coefficient precision \* lpc order).
 `Coded residual` | Encoded residual
 
-See [section on Constant subframe](#constant-subframe) on how the warm-up samples are stored unencoded. The unencoded predictor coefficients are stored the same way as the warm-up samples, but the number of bits needed for each coefficient is defined by the predictor coefficient precision. While the prediction right shift is signed two's complement, this number MUST be positive.
+See [section on Constant subframe](#constant-subframe) on how the warm-up samples are stored unencoded. The unencoded predictor coefficients are stored the same way as the warm-up samples, but the number of bits needed for each coefficient is defined by the predictor coefficient precision. While the prediction right shift is signed two's complement, this number MUST NOT be negative, see section [past changes](#restriction-of-LPC shift-to-non-negative-values) for an explanation why this is.
 
 Please note that the order in which the predictor coefficients appear in the bitstream corresponds to which **past** sample they belong. In other words, the order of the predictor coefficients is opposite to the chronological order of the samples. So, the first predictor coefficient has to be multiplied with the sample directly before the sample that is being predicted, the second predictor coefficient has to be multiplied with the sample before that etc.
 

rfc_backmatter.md

@@ -77,6 +77,10 @@ Another change to the specification was deemed necessary during standardization
 
 One significant addition to the format was the residual coding method using a 5-bit Rice parameter. Prior to publication of this addition in July 2007, there was only one residual coding method specified, a partitioned Rice code with a 4-bit Rice parameter. The range offered by this proved too small when encoding 24-bit PCM, therefore a second residual coding method was specified identical to the first but with a 5-bit Rice parameter.
 
+## Restriction of LPC shift to non-negative values
+
+As stated in section [linear predictor subframe](#linear-predictor-subframe), the predictor right shift is a number signed two's complement, which MUST NOT be negative. This is because right shifting a number by a negative amount is undefined behavior in the C programming language standard. The intended behavior was that a positive number would be a right shift and a negative number a left shift. The FLAC reference encoder was changed in 2007 to not generate LPC subframes with a negative predictor right shift, as it turned out that the use of such subframes would only very rarely provide any benefit and the decoders that were already widely in use at that point were not able to handle such subframes.
+
 # Interoperability considerations
 
 As documented in appendix [past format changes](#past-format-changes), there have been some changes and additions to the FLAC format. Additionally, implementation of certain features of the FLAC format took many years, meaning early decoder implementations could not be tested against files with these features. Finally, many lower-quality FLAC decoders only implement a subset of FLAC features required for playback of the most common FLAC files.