On 7/31/06, Derek Scherger <[EMAIL PROTECTED]> wrote:
If it helps, the restrictions stuff only uses app.lua (for the ignore_file hook) and the app.depth option setting. It doesn't touch anything else in app_state so there's really no need to pass the whole thing in.
I saw that, but what with restrictions being used all over the code, I thought it would be unwise to add yet more refactoring to the review load for this branch. Also, it makes even less sense to provide app.depth as an argument to update_current_roster_from_filesystem. An alternative, I suppose, is to remove the ucrff overload that doesn't take a node_restriction argument, and make all callers provide one. This would percolate to maybe_update_inodeprints' callers, too. Most callers are commands, so they have visibility of app_state already. Another alternative is to drop the use of node_restriction entirely from ucrff, make it always scan everything, but this might have bad performance consequences. The other thing app gets used for is app.prog_name, which arguably belongs in the global ui object instead... zw zw _______________________________________________ Monotone-devel mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/monotone-devel
