Danek Duvall wrote:
Brock Pytlik wrote:
http://cr.opensolaris.org/~bpytlik/ips-11104-v1/
Is there a chance that rebuild_index_from_scratch() could raise an
exception? If so, then presumably the clears should be un-done. Or move
them after the index rebuild.
I don't understand this. If rebuild_index from scratch raises an error,
then the indexes will not be written to disk, so undoing the clears
won't make any difference.
Should the test at the top be on self._data_fast_remove as well?
If you mean the test in indexer to decide whether to do the rebuild from
scratch, then no, it shouldn't. Having things in fast_add means we have
to go and read all those manifests entirely from disk. Having things in
fast_remove simply means we discard the data we read from disk. I
suppose it is possible that if you removed 50% of your packages, that
the regular indexes would be so substantially smaller that it would be
worth rebuilding the indexes. In short, having fast_add be large is a
substantial penalty, I see fast_remove as having much less impact.
Does that make things clearer?
Brock
Danek
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
