Values begin, end variables in pns.c may not be correct in the case of short window sequences

[atimmernans]

Some months the variables offs, size where replaced by begin and end variables which are basically calculated as:

min(group* (frame_length/8) + ics->swb_offset[x], ics->swb_offset_max)

In the case of HEIGHT_SHORT_SEQUENCE, ics->swb_offset_max is frame_length/8 (see specrec.c) which means for group > 0 the values of both begin and end will always be equal to frame_length/8. => this means the subsequent calls to generate_rand_vector generate nothing for group > 0.

Speaking of the ics->swb_offset table, it can be safely replaced by a pointer to the relevant constant table in specrec.c. The code in there copies into the swb_offset table the entries of the relevant swb_offset_xxx_yy table and put frame_length or frame_length/8 in swb_offset[ics->num_swbs] but as you can see swb_offset_xxx_yy[ics->num_swbs] already contains the appropriate value, to you can just point to the table directly instead of taking a copy.

This also points to the fact that all those min(ics->swb_offset[x], ics->swb_offset_max) in various source files are unnecessary. Since ics->swb_offset[ics->num_swbs] = ics->swb_offset_max this means that ics->swb_offset[x] is always <= ics->swb_offset_max as long as x <= ics->num_swbs (i.e. you don't read past the table end). This in turn is guaranteed as long as ics->max_sfb <= ics->num_swbs.