fix Track0 sector size in formatting Hard Drive.

It should not be different from the sector size
on other tracks.

Also a unit test that breaks when the bug exists.

FIXES #28

Author: strickyak

os9/os9format.c

@@ -274,6 +274,8 @@ int os9format(int argc, char **argv)
 				i = 0;
 			}
 		}
+
+		sectorsTrack0 = sectorsPerTrack;
 	}
 
 	/* walk command line for pathnames */

unittest/librbftest.c

@@ -16,6 +16,7 @@ void test_os9_format()
 	// test format of a non-existent disk image
 	unsigned int totalSectors, totalBytes;
 	int clusterSize = 0;
+
 	ec = _os9_format("test.dsk", 0, 35, 18, 18, 1, 256, &clusterSize, "Test Disk", 8, 8,
 			 1, 1, 0, 0, &totalSectors, &totalBytes);
 	ASSERT_EQUALS(0, ec);
@@ -27,6 +28,65 @@ void test_os9_format()
 			 8, 8, 1, 1, 0, 0, &totalSectors, &totalBytes);
 	ASSERT_EQUALS(0, ec);
 
+
+	// Test format of a hard drive with 65000 sectors.
+	// Getting the parameters for _os9_format:
+	//   $ gdb os9/os9
+	//   (gdb) b _os9_format
+	//   (gdb) run format test.dsk -l65000
+	// Breakpoint 1, _os9_format (pathname=0x7fffffffe0ad "test.dsk", os968k=0,
+	// tracks=125, sectorsPerTrack=4, sectorsTrack0=4, heads=130, sectorSize=256, 
+	// clusterSize=0x7fffffffdae8, diskName=0x5555555761df "CoCo Disk",
+	// sectorAllocationSize=8, tpi=0, density=0, formatEntire=0, isDragon=0, isHDD=1, 
+	// totalSectors=0x7fffffffdaec, totalBytes=0x7fffffffdaf0) at ../../../librbf/librbfformat.c:25
+	//   (gdb) cont
+	// Continuing.
+	// Format Summary
+	// --------------
+	// Geometry Data:
+	//       Cylinders: 125
+	//           Heads: 130
+	//   Sectors/track: 4
+	//     Sector size: 256
+	//
+	// Logical Data:
+	//   Total sectors: 65000
+	//   Size in bytes: 16640000
+	//    Cluster size: 1
+
+	ec = _os9_format("test.dsk", /*68k=*/0,
+			/*tracks=*/125, /*sPT=*/4, /*sT0=*/4, /*heads=*/130, /*sSize=*/256, &clusterSize,
+			 "HD-Test",
+			 /*sas=*/8, /*tpi=*/0, /*density=*/0, /*fE=*/1, /*isD=*/0, /*isHDD=*/1,
+			 &totalSectors, &totalBytes);
+	ASSERT_EQUALS(0, ec);
+	ASSERT_EQUALS(65000, totalSectors);
+	ASSERT_EQUALS(65000 * 256, totalBytes);
+
+#ifdef unix
+	// In order to demonstrate the bug described in
+	//     https://github.com/nitros9project/toolshed/issues/28
+	// we have to execute the `os9 format` command.
+	// The bug is not in _os9_format, but in preparing its parameters.
+	ec = system("rm -f test.dsk; ../os9/os9 format test.dsk -l65000");
+	ASSERT_EQUALS(0, ec);
+	// When `os9 format` was broken, it would succeed, but later
+	// operations would fail, like `os9 makdir`:
+	ec = _os9_makdir("test.dsk,/CMDS");
+	ASSERT_EQUALS(0, ec);
+	/*
+		// ... But if `os9format` were in the tinytest linkage, this could work,
+		// instead of using system():
+		char* argv[] = {
+			"os9", "format", "test.dsk", "-l65000", NULL
+		};
+		ec = os9format(4, argv);
+		ASSERT_EQUALS(0, ec);
+		ec = _os9_makdir("test.dsk,/CMDS");
+		ASSERT_EQUALS(0, ec);
+	*/
+#endif
+
 	// TODO: test format with oddball parameters to make sure it can survive
 }