On 23 March 2016 at 20:39, <[email protected]> wrote: > Author: kotkov > Date: Wed Mar 23 17:39:36 2016 > New Revision: 1736357 > > URL: http://svn.apache.org/viewvc?rev=1736357&view=rev > Log: > Introduce `--no-flush-to-disk' option for `svnadmin load'. > > The option can be used to to dramatically speed up the load process when > there's no need to ensure that the resulting data survives a system crash > or power loss — e.g., when loading a dump into a fresh new repository. > > This is one of the ideas in http://svn.haxx.se/dev/archive-2015-09/0187.shtml > (Subject: "Whiteboard -- topics list on the white board"). > Nice option, see my review inline.
[...] > Modified: subversion/trunk/subversion/libsvn_fs_fs/fs.c > URL: > http://svn.apache.org/viewvc/subversion/trunk/subversion/libsvn_fs_fs/fs.c?rev=1736357&r1=1736356&r2=1736357&view=diff > ============================================================================== > --- subversion/trunk/subversion/libsvn_fs_fs/fs.c (original) > +++ subversion/trunk/subversion/libsvn_fs_fs/fs.c Wed Mar 23 17:39:36 2016 > @@ -292,6 +292,7 @@ initialize_fs_struct(svn_fs_t *fs) > fs_fs_data_t *ffd = apr_pcalloc(fs->pool, sizeof(*ffd)); > ffd->use_log_addressing = FALSE; > ffd->revprop_prefix = 0; > + ffd->flush_to_disk = FALSE; I think would be more safe to initialize FLUSH_TO_DISK to TRUE to match old behavior and prevent possible mistakes in future when some other codepath will initialize_fs_struct() without calling read_global_config(). [...] > Modified: subversion/trunk/subversion/libsvn_fs_fs/revprops.c > URL: > http://svn.apache.org/viewvc/subversion/trunk/subversion/libsvn_fs_fs/revprops.c?rev=1736357&r1=1736356&r2=1736357&view=diff > ============================================================================== > --- subversion/trunk/subversion/libsvn_fs_fs/revprops.c (original) > +++ subversion/trunk/subversion/libsvn_fs_fs/revprops.c Wed Mar 23 17:39:36 > 2016 > @@ -776,7 +776,7 @@ switch_to_new_revprop(svn_fs_t *fs, > apr_pool_t *pool) > { > SVN_ERR(svn_fs_fs__move_into_place(tmp_path, final_path, perms_reference, > - pool)); > + TRUE, pool)); Why ffd->flush_to_disk() is not used here? As far I remember revprops-only loading from dump is one of the use case and we already have optimized codepath for them. -- Ivan Zhakov
