"port atlas not found"

[Ryan Schmidt]

> On Dec 4, 2011, at 17:01, Jeremy Lavergne wrote:
> 
>>> I have seen those conversations. I don't know how to explain them.
>> 
>> I'm beginning to suspect our portindex isn't create anatomically. Can we verify that it is?

On Dec 4, 2011, at 17:02, Jeremy Lavergne wrote:

>> anatomically
> 
> Atomically, even.

Atomically, meaning all or nothing? Correct, portindex does not function that way. If any individual port fails to index, the other ports still get indexed. This is intentional, I'm sure; we wouldn't want a syntax error in a single portfile to prevent all other ports from indexing. Of course since some time now we don't rebuild the entire index each time; we just add new and changed ports to the index. I am not sure what MacPorts does now when a port is deleted. (Does it get deleted from the index?) It seems to me like it might still be occasionally beneficial to delete the PortIndex and let it be regenerated from scratch. This describes the experience of using svn to fetch your ports tree; for most users, who use rsync, they get the pre-generated PortIndex on the rsync server, so their index should always be fresh (if we assume the rsync server's index is accurate).