Ports with a wrong PKGORIGIN

[Chuck Swiger]

Oliver Eikemeier wrote:
> Pav Lucistnik wrote:
>> Suggestion to eik: what about mailing ports-developers@ instead of
>> ports@? I read ports-developers immediately, but ports@ only once in
>> a while.
> 
> I'm happy with that, since only ports committers can do something to
> fix this anyway. OTOH, the same rationale that are valid for a
> broken INDEX apply to PKGORIGIN and PKGVERSION: Port users might want
> to know that the will run into trouble when they are trying to
> update the port, and CCing ports-developers seems overambitious.

While I am concerned that the ports infrastructure be in a consistent state, I 
cannot say that seeing broken INDEX or these "wrong PKGORIGIN" messages is 
especially useful.

The main value of the ports system to me is that generally two people (a 
port's maintainer and a ports committer) have worked on a particular piece of 
software to try to make sure the software runs well on FreeBSD and follows the 
conventions (following the hier manpage, being installable and deinstallable 
via the pkg_* tools, respecting compiler flags, etc).

If there is a problem with an update to a port, most of the time, I am 
perfectly willing to wait for those people to sort things out.  In a few 
cases, perhaps I might make my own local changes to coerce things to proceed, 
but for the most part, I think these alerts could be sent to a smaller, more 
targeted audience that either has a commit bit to fix the issue and/or is a 
maintainer of the software.

Also, if the ports infrastructure has scripts which can verify whether the 
"referential integrity constraints" are broken by a commit (to snarf a 
database term), it might be more useful to use such scripts as part of the CVS 
pre-commit checks.  Of course, sometimes that kind of checking can make it a 
real pain to make a large change in steps where the intermediate phases are 
inconsistent.  However, the use of some override mechanism-- perhaps checking 
whether the change is done by a portmgr, or has an "Approved by: portmgr" in 
the CVS commit message, would be a good solution.

-- 
-Chuck