[kris@FreeBSD.org: cvs commit: ports/lang/nawk Makefile]

horio shoichi bugsgrief at bugsgrief.net
Wed Feb 25 07:24:58 PST 2004


On Tue, 24 Feb 2004 18:03:25 -0800
Kris Kennaway <kris at obsecurity.org> wrote:
> 
> No, the port is broken for anyone who fetches a fresh copy of it.  The
> BROKEN tag merely documents this: the port must either be fixed to use
> the new distfile or changed to fetch an old version of it.  That's the
> way we do things around here :)
> 
> Kris
> 

Marking as BROKEN has far stronger meaning than mere documenting.
It makes others retrieve the port hard. Apparently it's worse than
allow using only three weeks older source. If you are not encouraging
people disregard your "document" you know a lot of better ways.

If something is broken, it is not the individual port. It should be
beyond each port's expectation that the first failure does not recover.
And it is also beyond our expectation. The brokenness is the separation
of fetch target that handle round trip among sites and checksum target
that does md5 calculation. With current mechanism the trip never goes
beyond first successful (in the sense of transmission) download. In
other word, if the transmission itself (i.e., FETCH_CMD) does md5 check
as part of its state then fetch target will just happily make the trip*.

	*One would argue fetch and checksum are categorically different
	things. Suffice to say fetch and checksum targets have been
	trying to establish feedback among the two for years but it
	looks rather logical to handle the two as unrelated independent
	subjects.

So, as for downloading capability, I'm able to download older version
of awk.tar.gz with make checksum command, although the first attempt
from master site fails with md5 sum error. To me the port is not
broken since my script, given as FETCH_CMD, returns md5 result to fetch
target which in turn is able to trip to the next site.


Thus, the port is not broken in theory and practice.


horio shoichi



More information about the freebsd-ports mailing list