atom feed6 messages in org.freebsd.freebsd-portsRe: requesting enhancements to addpor...
FromSent OnAttachments
Mikhail TeterinApr 18, 2001 6:08 am 
Will AndrewsApr 18, 2001 8:23 am 
mi...@aldan.algebra.comApr 18, 2001 9:18 am 
Maxim SobolevApr 18, 2001 9:52 am 
mi...@aldan.algebra.comApr 18, 2001 10:24 am 
Will AndrewsApr 18, 2001 1:29 pm 
Subject:Re: requesting enhancements to addport :)
From:Will Andrews (wi@physics.purdue.edu)
Date:Apr 18, 2001 1:29:40 pm
List:org.freebsd.freebsd-ports

On Wed, Apr 18, 2001 at 07:52:28PM +0300, Maxim Sobolev wrote:

Well, gziped CVSROOT/modules currently is 55KB long (transfer size is 110KB, because file have to be transferred to and then from client), so dialup committers would kill you.... In addition, it would greatly increase possibility of stale lock on CVSROOT/modules (Bad Thing, BTW) due to dualup going down in the middle of transfer.

Multiplied by however many times someone wants to commit new ports. When I wrote addport, I was still behind a dialup link, and I greatly understood how much it would impede new ports. It's simply not feasible to do this sort of thing without enough bandwidth.

There is an option to not commit CVSROOT/modules delta; people can write a quick script that adds it for themselves. I chose to take this route once -- January 4, 2001, where I committed 79 perl5 module ports in the space of 90 minutes. This sort of thing would not be possible for dialup users because of the load of CVSROOT/modules.

In addition, I have tried really hard to make addport scalable to multiple new ports. I still haven't rewritten /usr/local/bin/modulesupdate to accomodate multiple new modules at once, so that's really the first place to start.

I'm objecting to this. In addition to a valid arguments supplied by the Will and my arguments above in your scheme cvs history of CVSROOT/modules will be bloated with unnecessary description of each port added.

Yup.