On Sunday 06 February 2011 20:03:13 Cedric Sodhi wrote: > On Sun, Feb 06, 2011 at 05:54:19PM +0000, Neil Bothwick wrote: > > On Sun, 6 Feb 2011 12:53:20 +0100, Cedric Sodhi wrote: > > > > > 1. With a sudden change portage would simply resync to a new > > > > > directory, the old tree would rot in /usr > > > > > > > > And people would hit problems because /var bas filled up! I'm not > > > > saying the current default is right, it's not, but you are > > > > over-simplifying the work involved in making a change. > > > > > > I disagree. You are overcomplifying it instead. The proposed patch > > > would involve exactly: > > > > > > 1.) Change the default (the value that is used if no explicit value is > > > given) > > > > > > 2.) etc-update make.conf to explicitly specify the old location as the > > > desired value. > > > > > > Period. > > > > So now you've added another step not previously mentioned, but one that > > just happens to answer the point I made? Your previous 1 statement is > > now no longer true, portage would not resync to a new directory, and the > > old one in /usr would continue to be used, only new installs would be > > affected. > > Correct.
It think this proposal (to only change portage for new installs) is eminently doable, with enough early e-warnings about it and changes in docs. It could be introduced with a change in the make.profile and require explicit user intervention. I'd vote for it - if anyone is counting. > > > Your attempts to argue that patching portage with that simple change > > > would introduce problems of unpreceeded magnitude are pharisaic. It's > > > the same though significantly simpler as other updates to whatever > > > package you like. I don't think Neil's is being pharisaic in his statement. Some machines may need repartitioning to make portage fit in /var. I've got at least one old box where this would apply and will want to keep portage in /usr until I have time to shift things around. It's not as simple as e.g. recreating your xorg.conf to get yout X back, or disabling hal to find your mouse again. Without adequate early warnings and suitable advice a good number of users will complain that their machines are borked and blame the devs. That's why I'm suggesting that manual intervention should work as a check that the user knows what their doing. Packages that barf if /usr/portage is changed (don't know of any) will need to be managed via bug reports, or hopefully their devs will be clued enough to head this off at the pass. Anyway, just my 2c's. -- Regards, Mick
signature.asc
Description: This is a digitally signed message part.
