Hi, > >> 2. Patches that divide gpt(8) code into two parts: core functions > >> (backend) and /sbin/gpt (frontend). That's will be useful for code share; > >> 3. (probably, if time will allow) Patches to gpt(8), fdisk(8), > >> disklabel(8), sunlabel(8) and sysinst to share code into one library that > >> contains all needed partitioning functions; > > That'd be very nice imho. I just wonder if having a library for that is > > more difficult than just having the tool usage itself unified. > > No, it is not more difficult if done at least minimally carefully. > In most cases main() should consist only of argument parsing and calling > approporiate functions. Thus it is should be indifferent whether you > write a library or a standalone program. (Making it a library even makes > unit testing simpler, to appease those who drank too much of testing kool > aid.) yep, right. I had something else in mind, namely to have a single interface which decides whether to use mbr, gpt or disklabel partitioning (or whatever else is out there) depending on an argument. That would in turn require actually recoding stuff. > >> * Implementing ability to install pkgsrc from sysinst; > > This was already implemented, though not committed. See the appropriate PR > > for that (this would fix binary package installation as well - or, > > somehow). > > What does "install pkgsrc" mean here? > How do you propose to handle the difference in release cycles? That was done here: http://mail-index.netbsd.org/tech-install/2012/01/23/msg000223.html I think always installing the most recent stable version would be right, as that is what most users do anyway. As you're asked for the path, you could as well choose current if you know its path, or another stable release. Regards, Julian
Attachment:
signature.asc
Description: PGP signature