pkgsrc build from source issue?

[Jeremy C. Reed]

In reply to http://leaf.dragonflybsd.org/mailarchive/users/2006-01/msg00026.html

On Thu, 5 Jan 2006, Matthew Dillon said:

>   It's a bit of a hack.  I agree we need to better-integrate the pkgsrc
>   tools.  I am considering bringing in just the bootstrap and related
>   required tools into our CVS tree for just that purpose.  We would
>   keep them 100% synchronized, but it would remove the need to separately
>   fetch the pkgsrc bootstrap and related packages.

For the future, maybe we can work on getting it so the bootstrap is not 
needed.

I heard someone was considering working on merging in support needed from 
DragonFly into bmake or visa versa.

And the ftp(1) is already tnftp (I don't know how up-to-date).

And digest is simple to include to DragonFly: as itself, or as a wrapper 
around openssl or sha1 and md5.

I haven't looked to see how far pax is diverged or is different. Maybe 
DragonFly's pax is good enough. (On an unrelated note, NetBSD's pax has an 
awesome feature where it can use a mtree(8) `specfile' specification when 
building an archive.)

As for libnbcompat, I assume DragonFly is close enough to exclude that 
too. (Maybe I'll try to build pkg_install just natively on DragonFly 
without it.)

As for mtree: it is not needed for pkgsrc. In fact, I use NO_MTREE on some 
systems. (Anyways, DragonFly's mtree is used already.)

And DragonFly's sed is good enough and is already used.

And a BSD install(1) is not needed.

In other words, maybe the bootstrap is not needed other than just pkg_* 
tools (and make some .mk files).

(Please CC me on replies.)

 Jeremy C. Reed

 	  	 	 BSD News, BSD tutorials, BSD links
	  	 	 http://www.bsdnewsletter.com/