portmanager endlessly looping in x11

RW rwmaillists at googlemail.com
Thu Sep 9 01:19:55 UTC 2010

On Wed, 08 Sep 2010 17:41:32 -0400
Chuck Robey <chuckr at telenix.org> wrote:

> I'm making no mistake tho, moved from
> portmanager to portmaster) which doesn't seem to have this uneveness,
> so while it takes a whole lot longer to work than portmanager (it
> uses slow but sure shell utils for it's databases)

I don't know why people think  portmanager is fast. It may be written
in C, but in most  upgrades it builds *many* more ports than portmaster
or portupgrade would do. The point of a portmanager is to upgrade
correctly with as little human intervention as is possible; and it
sacrifices a lot of cpu cycles to that end.

> it really does a far more reliable job of things.  

Something it does, sometimes it doesn't. portmanager will handle a lot 
upgrades correctly even when instructions in UPDATING that are
required for portmaster are ignored. With portupgrade I've had a few
problems where the UPDATING entry was made after I updated the ports.
These have been much more serious than harmless looping.

> One really big irritation was
> how portmanager would rebuild something completely successfully 3
> times, but since it would fail its dependency scans, it never would
> recognize that any of those looping apps had been rebuilt.  Very
> puzzling, until I realized about the dependency problems.

I didn't follow what you were were saying, but portmanager has at least
two features that can lead to looping. One is that it can rebuild ports
when it detects a dynamic change in port dependencies. The other is
when it tries to resolve conflicts by package deletion and iterative
rebuilding. There are probably more. AFAIK it shouldn't actually loop
endlessly because of its "three-strikes database".

In my experience portmanager goes through good and bad patches
as the installed ports change and their dependencies evolve. Even when
it's not completing successfully by it's own criteria, it usually does
the job without leaving any real problems of its own making. 

More information about the freebsd-ports mailing list