Int 21.7303 returns one too many in count of total clusters

[ecm-pushbx]

This is where function 7303h uses dpb_size or dpb_xsize:

kernel/kernel/dosfns.c

Line 948 in 5de2eb1

(ISFAT32(dpbp) ? dpbp->dpb_xsize : dpbp->dpb_size);

But function 36h uses dpb_xsize minus one:

kernel/kernel/dosfns.c

Line 835 in 5de2eb1

ntotal = dpbp->dpb_xsize - 1;

Or dpb_size minus 1 on FAT12/FAT16:

kernel/kernel/dosfns.c

Line 859 in 5de2eb1

*nc = dpbp->dpb_size - 1;

The field is marked as "number of clusters plus 1":

kernel/hdr/dcb.h

Line 49 in 5de2eb1

UWORD dpb_size; /* # of clusters+1 on media */

In other words the DPB field gives the maximum valid cluster number (0 and 1 are reserved), plus one gives amount of FAT entries, plus one minus two = minus one gives amount of data clusters.

[andrewbird]

I updated the test for int21/0x7303 on Dosemu's test suite to check the actual cluster counts with a FAT32 image, previously I'd been checking that we were within 5% of expected values. I can confirm that with MS-DOS 7.1 the cluster counts agree with linux dosfsck, and that with both current git kernels of FreeDOS and FDPP they are off by one. @boeckmann I see you have a fix, will you be raising a PR?

[boeckmann]

@boeckmann I see you have a fix, will you be raising a PR?

I just did it :)