On Thu, 29 Nov 2007, David Glasser wrote:
> So 'svn log URL' (without --stop-on-copy) does this thing where it
> traces the path back across history, so not all of the log entries
> actually contain the path the user typed on the command line.
>
> log -g ignores that.
>
> Look at libsvn_repos/log.c(do_logs). It does all this fancy stuff to
> calculate an array of path_info structs (histories). But then it just
> passes plain old 'paths' to send_logs, which tries to calculate
> mergeinfo based on the paths the user typed.
>
> This is wrong, and on my branch which actually checks for path
> existence for the FS mergeinfo queries, throws an error. (But the
> problem is that it is *wrong*, not that it throws an error.)
We need 'log -g' to intermix behavior like whatever Mike's
get_location_segments() is doing at the libsvn_repos level to re-introduce
the history tracing. Where should we inject that history tracing, and how
will it interact with these path_info structs?
--
Daniel Rall
- application/pgp-signature attachment: stored
Received on Sun Dec 2 01:17:47 2007