portupgrade O(n^m)?
Olivier Warin
daffy at xview.net
Wed Feb 14 19:20:09 UTC 2007
daffy at katia:~ %> pkg_info | wc -l
-[19:49]-
917
Really portupgrade becomes clearly not so usable for me after I
switch to Xorg 7.2RC which install 300 more packages, my workstation
is a xSeries 226 with a Xeon 2,8Ghz 1Go DDR2. So I can imagine what
it does on a laptop...
This issue is not only related to portupgrade, pkg_add a new port
takes far too long now... and make index each time I upgrade my ports
is awfull too.
"Someone has to do something" (tm)
Regards,
Le 14 févr. 07 à 19:36, Coleman Kane a écrit :
> On 2/14/07, John Nielsen <lists at jnielsen.net> wrote:
>>
>> On Wednesday 14 February 2007 12:41, David Gilbert wrote:
>> > I have 734 ports installed on my laptop right now. I'm pretty
>> sure,
>> > at times, I've had over 1000 ports on my laptop.
>> >
>> > On machine with moderate numbers of ports (most servers seem to
>> have
>> > 50 to 200 ports), portupgrade takes a moderate amount of time to
>> start
>> > work. On machines like my laptop, portupgrade seems to take
>> much more
>> > time to run. I assume it's solving the dependency graph before it
>> > decides what to upgrade first, but is this truly a O(n^2)
>> problem? It
>> > seems like the implemented algorithm is O(n^2).
>>
>> Just a "me too". I noticed a huge increase in time for portupgrade
>> when I
>> started using the modular Xorg ports tree and upgraded to X.org
>> 7.2RC. The
>> number of installed ports on my machine went from just over 300 to
>> well
>> over
>> 600 as a result of the upgrade. Specifying small numbers of ports
>> (without
>> globbing) to portupgrade doesn't seem to take much more time,
>> but "portupgrade -a" or anything similar takes forever now. If
>> there is an
>> optimization to be made there it would be good to do it before
>> modular
>> xorg
>> hits the official tree.
>>
>> JN
>
>
> I've also had this problem. I have found that if I perform a
> "portsdb -U &&
> pkgdb -F" every time following a cvsup that portupgrade doesn't try
> to go
> through the full ports indexing steps again.
>
> It is still slow, and any improvement that can be made should be.
> It is
> already a significant enough pain that most ports build in a
> shorter amount
> of time than it takes portupgrade to update its database.
>
> --
> Coleman Kane
> _______________________________________________
> freebsd-hackers at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
> To unsubscribe, send any mail to "freebsd-hackers-
> unsubscribe at freebsd.org"
--
Olivier Warin
More information about the freebsd-hackers
mailing list