On 19/01/11 11:48, Dan McGee wrote:
On Tue, Jan 18, 2011 at 7:41 PM, Allan McRae<[email protected]> wrote:
On 19/01/11 10:50, Dan McGee wrote:
Nothing will use this yet, but we might as well read the info in while
processing the database. The drawback would be vastly increased memory
usage, so we might want to add a flag enabling/disabling the loading of
this
data.
This will at least for now be unused by pacman, but a pkgfile.c program at
least now has the option of using standard libalpm interfaces to search
for
file ownership.
Signed-off-by: Dan McGee<[email protected]>
---
Thoughts from the peanut gallery?
Not only will memory usage increase massively, but the read speed of the
sync databases will also take a decent hit. I'd very much want to
reintroduce some lazy loading to the sync database reading just for this,
especially since that files list would only be needed for a -So operation.
Sure- but I wasn't planning on telling anyone to point pacman at any
databases with files anytime soon.
Well, the feature is now in repo-add so I am sure there will be some
repos that start using it... As long as they are not repos that I am
using, all is good!
However, thi would be extremely
useful for a libalpm-based pkgfile replacement as I said in the commit
message. Right now, there is no possibility at all of doing that with
libalpm.
I'm also leaning against reintroducing lazy loading, and more toward a
DB-register-time argument that will allow or deny the loading of files
entries from the sync DB handed to the library. This will keep us from
having to add back all the cache BS we worked hard to remove.
Indeed, that does sound a better idea.
Allan
