In this case, if CODING->src_pos is positive, it is a position of
the source text in the buffer, otherwise, the source text is in the
gap area of the buffer, and CODING->src_pos specifies the offset of
- the text from GPT (which must be the same as PT). If this is the
- same buffer as CODING->dst_object, CODING->src_pos must be
- negative.
+ the text from the end of the gap (and GPT must be equal to PT).
+
+ When the text is taken from the gap, it can't be at the beginning
+ of the gap because the new decoded text is progressively acumulated
+ at the beginning of the gap before it gets inserted at PT (this way,
+ as the output grows, the input shrinks, so we only need to allocate
+ enough space for `max(IN, OUT)` instead of `IN + OUT`).
If CODING->src_object is a string, CODING->src_pos is an index to
that string.