Hi Jim, > 1. There is a check on the descriptor returned by sane_get_select_fd() > which always fails even when my backend has returned a valid small > integer file descriptor. The check is implemented like > "check(ERR,(fd <0), ......". I think it should be > "check(ERR,(fd>=0),...."
That's a tstbackend bug. Your backend seems to be the first tested one to support sane_get_select_fd. > > 2. In the scanning tests tstbackend starts a scan and then toggles > the state of the io_mode eventually leaving the fd in non-blocking > mode. It then attempts to do a 1 byte read. Due to being very slow > scanner, my DS3000 backend has no data ready on the first call to > sane_read() and therefore returns SANE_STATUS_GOOD and sets the length > to 0 as suggested by the SANE API doc. tstbackend thinks this is an > error and aborts the rest of the scan. I would expect that either > tstbackend should use the descriptor in blocking mode or should accept > return value of SANE_STATUS_GOOD with a returned length of 0 bytes? > Ditto with sane_set_io_mode. I'll fix that. Frank.
