[pacman-dev] lstat issue / fileconflict001 pactest + new failing pactest

[Nagy Gabor]

> OK. My main problems with rollback:
> 1. this doesn't look easy to implement at all
> 2. in case of powerbreak (or any small hidden bugs) I can get MBs of
> "spam" on my HDD.
I insert 3rd and 4th points here:
3. User does a pacman -Su, which will start extracting about ~1GB data.
After 98% of the process pacman detects a fileconflict; and it rollbacks
the transaction... User won't be happy. He repeat -Su, and he get an
_other_ fileconflict error at about 99%. User is quite angry now ...
a.) So, you should probably keep the current negligent fileconflict
checking to collect (negligently) _all_ the fileconflicts. Which is not
really nice (and slow).
b.) Or you should continue(?!) the broken transaction to
detect as many bugs as possibly.
c.) "continue" option; grr
4. See the previous attached pactest file: correct rollback needs some
kind of universal transaction: currently our transactions often create
new pseudo transactions and this can mess things up...

Bye, ngaba