I've posted a new webrev:
http://cr.opensolaris.org/~dduvall/pkg-newlist2/
And the incremental from the previous:
http://cr.opensolaris.org/~dduvall/pkg-newlist1-2/
> > 920 + self.cache_catalogs()
> > 921 +
> > 894 922 if failed:
> > 895 923 raise RuntimeError, (failed, total,
> > succeeded)
> >
> > Is 920 and 921 intentionally before 922/923? Is the thought that we
> > should cache what we can? Is there any concern that if failed was true
> > that we would be caching an inconsistent or bad catalog state?
>
> That was the thought, but it's possible to get a bad catalog. I'll take
> another look at this.
So my thought here is that the catalog recv() methods should be more
careful to prevent bad data from hitting disk. Right now, they at least
collect all data from the network before writing anything to disk, but they
simply append to the existing file, rather than writing to a new one and
moving it into place. If we make that latter change, I don't think there's
any reason that having cache_catalogs() here will be problematic, while
moving it after the exception raising will mean that the successfully
downloaded catalogs won't be cached.
Thanks,
Danek
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
