long int ftell( struct _PDCLIB_file_t * stream )
{
- /* FIXME: A bit too fuzzy in the head now. stream->ungetidx should be in here
- somewhere.
+ /* If offset is too large for return type, report error instead of wrong
+ offset value.
+ FIXME: A bit too fuzzy in the head right now; stream->ungetidx should be
+ in here somewhere.
*/
if ( stream->pos.offset > ( LONG_MAX - stream->bufidx ) )
{
_PDCLIB_errno = _PDCLIB_EINVAL;
return -1;
}
- /* Position of start-of-buffer, plus:
- - buffered, unwritten content (for output streams), or
- - already-parsed content from buffer (for input streams)
+ /* ftell() must take into account:
+ - the actual *physical* offset of the file, i.e. the offset as recognized
+ by the operating system (and stored in stream->pos.offset); and
+ - any buffers held by PDCLib, which
+ - in case of unwritten buffers, count in *addition* to the offset; or
+ - in case of unprocessed pre-read buffers, count in *substraction* to
+ the offset. (Remember to count ungetidx into this number.)
+ Conveniently, the calculation ( ( bufend - bufidx ) + ungetidx ) results
+ in just the right number in both cases:
+ - in case of unwritten buffers, ( ( 0 - unwritten ) + 0 )
+ i.e. unwritten bytes as negative number
+ - in case of unprocessed pre-read, ( ( preread - processed ) + unget )
+ i.e. unprocessed bytes as positive number.
+ That is how the somewhat obscure return-value calculation works.
*/
- return (long int)( stream->pos.offset + stream->bufidx - stream->ungetidx );
+ return (long int)( stream->pos.offset - ( ( stream->bufend - stream->bufidx ) + stream->ungetidx ) );
}
#endif