re: CVS commit: src/sys

[christos%zoulas.com@localhost (Christos Zoulas)]

On Jan 3,  7:52am, mrg%eterna.com.au@localhost (matthew green) wrote:
-- Subject: re: CVS commit: src/sys

| this seems strange to me.
| 
| we have a published userland interface and you chose that
| one to change?  so userland apps must be fixed.  it seems
| much saner to me to fix the other in-kernel users to not
| conflict with the published interfaces.

There were 3 uses of DTYPE_ as far as disklabel constants go in the whole
source tree. Two of them were ifdef'ed on a particular architecture. They
were in:

bad144:
	#ifdef __i386__
		if (dp->d_type == DTYPE_SCSI)
			errx(1, "SCSI disks don't use bad144!");
		...

This looks like a horrible accident :-)

diskpart:
		if (dp->d_type == DTYPE_SMD && dp->d_flags & D_BADSECT && 
			    totsize == 0) {

disklabel:
	#ifdef VAX_ALTLABELS
		if (lab.d_type == DTYPE_SMD && lab.d_flags & D_BADSECT &&
			    lab.d_secsize == 512) { 

I would be hard pressed to find live SMD disks to do badsect manipulation -
and even if I could - I doubt that someone would want source compatibility
for bad sector management programs :-)

struct disklabel is on the way out; DTYPE_ constants referring to file
descriptors are a much more popular API even if they are supposed to be
kernel only constants...

christos