Please pass at least some path in the same filesystem as the resulting files. This allows detecting whether we can use a shorter sleep.
Bert Sent from Windows Mail From: Philip Martin Sent: Thursday, April 11, 2013 12:27 PM To: dev@subversion.apache.org julianf...@apache.org writes: > Author: julianfoad > Date: Tue Apr 2 19:53:43 2013 > New Revision: 1463721 > > URL: http://svn.apache.org/r1463721 > Log: > Fix some sleep-for-timestamps code: avoid missing some cases where we need > to sleep -- especially error cases -- and avoid sleeping in some cases where > we don't need to. Simplify the code by using a consistent pattern: make the > decision of whether to sleep at the lowest level possible (ideally in the WC > layer, but not yet), and do the sleep at the highest level possible > (immediately below the libsvn_client API). > * subversion/libsvn_client/copy.c > (do_wc_to_wc_copies_with_write_lock, > do_wc_to_wc_moves, > repos_to_wc_copy_single): Return a flag instead of sleeping here. A > multi-target repos-to-WC copy will now only sleep once, not once per > target. > (do_wc_to_wc_copies, > repos_to_wc_copy_locked, > repos_to_wc_copy, > try_copy): Pass the flag through. > (svn_client_copy6, > svn_client_move7): Handle the sleep here. > --- subversion/trunk/subversion/libsvn_client/copy.c (original) > +++ subversion/trunk/subversion/libsvn_client/copy.c Tue Apr 2 19:53:43 2013 > @@ -212,7 +214,6 @@ do_wc_to_wc_copies_with_write_lock(const > } > svn_pool_destroy(iterpool); > > - svn_io_sleep_for_timestamps(dst_parent, scratch_pool); > SVN_ERR(err); > return SVN_NO_ERROR; > } > @@ -2292,6 +2312,9 @@ svn_client_copy6(const apr_array_header_ > subpool); > } > > + if (timestamp_sleep) > + svn_io_sleep_for_timestamps(NULL, subpool); > + > svn_pool_destroy(subpool); > return svn_error_trace(err); > } Is there any reason you are passing NULL to svn_io_sleep_for_timestamps rather than dst_path? -- Certified & Supported Apache Subversion Downloads: http://www.wandisco.com/subversion/download
