You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Ben Collins <bc...@debian.org> on 2002/02/28 23:10:02 UTC
Updated [patch] new part 1 for log command
This is an updated patch. Basically it expands log usage to allow for
URL's and the -D option (which works with ra_local, but not ra_dav yet).
Updated man page aswell.
* libsvn_fs/tree.c(svn_fs_revisions_changed): Change to take an
array of stringbuf's for paths instead of one path. It will then
iterate to get a list of all revs for all paths (reduces amount
of trails for getting info on multiple paths).
* include/svn_fs.h: Updated decleration and comments for
svn_fs_revisions_changed.
* svnadmin/main.c(main): Updated use of svn_fs_revisions_changed
for lscr.
* tests/libsvn_fs/fs-test.c(revisions_changed): Likewise.
* libsvn_subr/sorts.c(svn_sort_compare_revisions): New function
for sorting an array of revisions.
* include/svn_sorts.h: Declare it.
* libsvn_repos/log.c(svn_repos_get_logs): Make use of newly
modified svn_fs_revisions_changed. We pass it the paths passed
to us (if any). The returned array of revs is then used to trim
down the logs to just those that the caller is intersted in.
* libsvn_client/log.c(svn_client_log): Enable the parsing and
usage of complete URL's.
* cmdline/main.c: Update usage info for log command. Also, allow
-D option for logs. This works with ra_local, but ra_dav doesn't
have the required function yet.
* man/svn.1(log): Update usage of log command.
--
.----------=======-=-======-=========-----------=====------------=-=-----.
/ Ben Collins -- Debian GNU/Linux -- WatchGuard.com \
` bcollins@debian.org -- Ben.Collins@watchguard.com '
`---=========------=======-------------=-=-----=-===-======-------=--=---'
Re: Updated [patch] new part 1 for log command
Posted by Ben Collins <bc...@debian.org>.
>
> > + base = ((svn_stringbuf_t **) (targets->elts))[0];
> > + svn_path_split (base, &basename, &file, pool);
>
> This split is only used by the nelts == 1 case (to get the file). It should
> move for clarity. Also, note that the nelts==1 case doesn't need basename,
> so you can pass NULL for that parameter (not &foo where foo is NULL).
No, file is only used in the == 1 case. basename is used in both.
> > + URL = svn_path_uri_encode(&path_str, pool);
>
> This is only needed for the nelts > 1 case, so it should be moved.
Again, it is needed for both. path_str is changed for the == 1 case, but
both cases need it (else where does the URL get set? :)
>
> The indentiation here looks weird. Tab characters? Please avoid the use of
> hard tabs in the code. In emacs, we ues (setq indent-tabs-mode nil)
>
I can setup vim to do this just fine :)
I'll clarify the above block with some comments on when and where each
variable gets used, and in which cases.
--
.----------=======-=-======-=========-----------=====------------=-=-----.
/ Ben Collins -- Debian GNU/Linux -- WatchGuard.com \
` bcollins@debian.org -- Ben.Collins@watchguard.com '
`---=========------=======-------------=-=-----=-===-======-------=--=---'
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Updated [patch] new part 1 for log command
Posted by Greg Stein <gs...@lyra.org>.
On Thu, Feb 28, 2002 at 06:10:02PM -0500, Ben Collins wrote:
>...
> +++ ./subversion/svnadmin/main.c Tue Feb 26 09:27:23 2002
> @@ -275,7 +275,7 @@
> {
> svn_revnum_t youngest_rev;
> svn_fs_root_t *rev_root;
> - apr_array_header_t *revs;
> + apr_array_header_t *revs, *paths;
> int i;
>
> if (argc != 4)
> @@ -284,14 +284,17 @@
> /* NOTREACHED */
> }
>
> + paths = apr_array_make (pool, 1, sizeof (svn_stringbuf_t *));
> + (*(svn_stringbuf_t **)apr_array_push(paths)) = svn_stringbuf_create(argv[3], pool);
See previous email re: stringbuf and the FS API.
>...
> +++ ./subversion/libsvn_fs/tree.c Wed Feb 27 22:05:48 2002
>...
> + /* Now sort the array and make it "go live". */
> + qsort(array->elts, array->nelts, array->elt_size,
> + svn_sort_compare_revisions);
Couldn't we end up with a lot of duplicates in there? Seems we should
probably remove dups.
>...
> @@ -2800,19 +2808,28 @@
> svn_error_t *
> svn_fs_revisions_changed (apr_array_header_t **revs,
> svn_fs_root_t *root,
> - const char *path,
> + const apr_array_header_t *paths,
> apr_pool_t *pool)
> {
> struct revisions_changed_args args;
> svn_fs_t *fs = svn_fs_root_fs (root);
> + int i;
> + svn_fs_id_t *tmp_id;
> + svn_stringbuf_t *this_path;
Generaly, we tend to put declarations into the "smallest scope possible".
Thus, the tmp_id and this_path variables could go into the for-loop block.
>...
> +++ ./subversion/libsvn_client/log.c Tue Feb 26 16:08:10 2002
> @@ -58,8 +58,8 @@
> svn_ra_plugin_t *ra_lib;
> void *ra_baton, *session;
> svn_stringbuf_t *URL;
> - svn_wc_entry_t *entry;
> - svn_stringbuf_t *basename;
> + svn_stringbuf_t *basename = NULL;
Why is this set to NULL? That doesn't seem to be needed. If so, then a
comment to clarify...
> + svn_string_t path_str;
> apr_array_header_t *condensed_targets;
> svn_revnum_t start_revnum, end_revnum;
>
> @@ -71,20 +71,68 @@
> "svn_client_log: caller failed to supply revision");
> }
>
> - SVN_ERR (svn_path_condense_targets (&basename, &condensed_targets,
> - targets, pool));
> + start_revnum = end_revnum = SVN_INVALID_REVNUM;
>
> - /* Get full URL from the common path, carefully. */
> - SVN_ERR (svn_wc_entry (&entry, basename, pool));
> - if (! entry)
> - return svn_error_createf
> - (SVN_ERR_UNVERSIONED_RESOURCE, 0, NULL, pool,
> - "svn_client_log: %s is not under revision control", basename->data);
> - if (! entry->url)
> - return svn_error_createf
> - (SVN_ERR_ENTRY_MISSING_URL, 0, NULL, pool,
> - "svn_client_log: entry '%s' has no URL", basename->data);
> - URL = svn_stringbuf_dup (entry->url, pool);
> + path_str.data = ((svn_stringbuf_t **)targets->elts)[0]->data;
> + path_str.len = ((svn_stringbuf_t **)targets->elts)[0]->len;
> +
> + /* Use the passed URL, if there is one. */
> + if (svn_path_is_url (&path_str))
> + {
> + svn_stringbuf_t *base, *file;
> +
> + /* The logic here is this: If we get passed one argument, we assume
> + * it is the full URL to a file/dir we want log info for. If we get
> + * a URL plus some paths, then we assume that the URL is the base,
> + * and that the paths passed are relative to it. */
I don't see where this logic occurs. There isn't a 'join' anywhere.
> + base = ((svn_stringbuf_t **) (targets->elts))[0];
> + svn_path_split (base, &basename, &file, pool);
This split is only used by the nelts == 1 case (to get the file). It should
move for clarity. Also, note that the nelts==1 case doesn't need basename,
so you can pass NULL for that parameter (not &foo where foo is NULL).
> + URL = svn_path_uri_encode(&path_str, pool);
This is only needed for the nelts > 1 case, so it should be moved.
>...
> +++ ./subversion/libsvn_repos/log.c Tue Feb 26 15:09:08 2002
> @@ -104,6 +104,8 @@
> apr_pool_t *subpool = svn_pool_create (pool);
> apr_hash_t *changed_paths = NULL;
> svn_fs_t *fs = repos->fs;
> + svn_fs_root_t *root;
> + apr_array_header_t *revs = NULL;
>
> if (start == SVN_INVALID_REVNUM)
> SVN_ERR (svn_fs_youngest_rev (&start, fs, pool));
> @@ -111,12 +113,32 @@
> if (end == SVN_INVALID_REVNUM)
> SVN_ERR (svn_fs_youngest_rev (&end, fs, pool));
>
> + if (paths && paths->nelts)
> + {
> + SVN_ERR (svn_fs_revision_root (&root, fs, (start > end) ? start : end, subpool));
> + SVN_ERR (svn_fs_revisions_changed (&revs, root, paths, subpool));
> + if (!revs || !revs->nelts)
> + return SVN_NO_ERROR;
The indentiation here looks weird. Tab characters? Please avoid the use of
hard tabs in the code. In emacs, we ues (setq indent-tabs-mode nil)
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Updated [patch] new part 1 for log command
Posted by Ben Collins <bc...@debian.org>.
> > + /* And put the revision into our array, if it doesn't already
> > + * exist. */
> > + for (t = 0; t < array->nelts && ! exists; t++)
> > + if (APR_ARRAY_IDX(array, t, svn_revnum_t) == revision)
> > + exists = 1;
>
> I'm mildly concerned about O(n^2) operation here, but I can't imagine this
> being all that large of an array. Regardless, when you find a match, you
> should 'break' out of the loop.
It does break out of the loop. Where it checks "! exist" in the for().
Once it finds a match, the loop will stop.
> An alternative to the nested loop is to remove dups after the qsort()
> occurs.
That's an idea. Means we only loop through once, recreating the array
and skipping duplicates.
> >...
> > + /* We split the path here. If this is the only target (a full path)
> > + we'll end up using BASENAME as the URL, and FILE as the path to
> > + search on. For both cases, BASENAME gets passed to
> > + svn_client__get_revision_number() below. */
> > + svn_path_split (base, &basename, &file, pool);
> >...
> > @@ -94,13 +149,14 @@
> > SVN_ERR (svn_client__open_ra_session (&session, ra_lib, URL, basename,
> > TRUE, TRUE, auth_baton, pool));
> >
> > + /* Get the revisions based on the users "hints". */
> > SVN_ERR (svn_client__get_revision_number
> > (&start_revnum, ra_lib, session, start, basename->data, pool));
> > SVN_ERR (svn_client__get_revision_number
> > (&end_revnum, ra_lib, session, end, basename->data, pool));
>
> When using a URL for an argument, there might not be a working copy. That
> implies that NULL should be passed for the 'base_dir' arg to open_ra_session
> (and FALSE for the do_store and use_admin flags). Also, NULL should be
> passed for the 'path' argument to get_revision_number.
Ok.
> >...
> > + if (revs)
> > + {
> > + int i, matched = 0;
> > + for (i = 0; i < revs->nelts && !matched; i++)
> > + if (this_rev == ((svn_revnum_t *)(revs->elts))[i])
> > + matched = 1;
>
> And break out of here, too.
Same as before, checks for !matched, so will break on the first match.
--
.----------=======-=-======-=========-----------=====------------=-=-----.
/ Ben Collins -- Debian GNU/Linux -- WatchGuard.com \
` bcollins@debian.org -- Ben.Collins@watchguard.com '
`---=========------=======-------------=-=-----=-===-======-------=--=---'
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Updated [patch] new part 1 for log command
Posted by Ben Collins <bc...@debian.org>.
On Fri, Mar 01, 2002 at 12:37:32AM -0500, Ben Collins wrote:
> * libsvn_fs/tree.c(svn_fs_revisions_changed): Change to take an
> array of stringbuf's for paths instead of one path. It will then
> iterate to get a list of all revs for all paths (reduces amount
> of trails for getting info on multiple paths).
Of course change this to "an array of `const char *'s" for paths"
--
.----------=======-=-======-=========-----------=====------------=-=-----.
/ Ben Collins -- Debian GNU/Linux -- WatchGuard.com \
` bcollins@debian.org -- Ben.Collins@watchguard.com '
`---=========------=======-------------=-=-----=-===-======-------=--=---'
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Updated [patch] new part 1 for log command
Posted by Ben Collins <bc...@debian.org>.
On Thu, Feb 28, 2002 at 08:43:49PM -0800, Greg Stein wrote:
> [ per IRC, there are still some problems in the svn_client_log() handling
> for URLs; Ben is going to redo that portion ]
>
> The patch looks good, but for the URL code in svn_client_log. I'd recommend
> returning SVN_ERR_UNSUPPORTED_FEATURE for that for now, and we'll get the
> rest of this in. The URL can be done in a second round. Thoughts?
Here's the new patch. I've implemented the suggestions (qsort, then
remove dups in svn_fs_revisions_changed). I've left the URL in place.
However, only the "URL PATH ..." case works, so trying to pass a URL
alone returns an unsupported error. Fixed the calls to
svn_client__open_ra_session and svn_client__get_revision_number in the
case of passed URL. Here's the changelog:
* libsvn_fs/tree.c(svn_fs_revisions_changed): Change to take an
array of stringbuf's for paths instead of one path. It will then
iterate to get a list of all revs for all paths (reduces amount
of trails for getting info on multiple paths).
* include/svn_fs.h: Updated decleration and comments for
svn_fs_revisions_changed.
* svnadmin/main.c(main): Updated use of svn_fs_revisions_changed
for lscr.
* tests/libsvn_fs/fs-test.c(revisions_changed): Likewise.
* libsvn_subr/sorts.c(svn_sort_compare_revisions): New function
for sorting an array of revisions.
* include/svn_sorts.h: Declare it.
* libsvn_repos/log.c(svn_repos_get_logs): Make use of newly
modified svn_fs_revisions_changed. We pass it the paths passed
to us (if any). The returned array of revs is then used to trim
down the logs to just those that the caller is intersted in.
* libsvn_client/log.c(svn_client_log): Enable the parsing and
usage of complete URL's. For now, passing a URL alone is
unsupported. User must pass a URL plus paths.
* cmdline/main.c: Update usage info for log command. Also, allow
-D option for logs. This works with ra_local, but ra_dav doesn't
have the required function yet.
* man/svn.1(log): Update usage of log command.
--
.----------=======-=-======-=========-----------=====------------=-=-----.
/ Ben Collins -- Debian GNU/Linux -- WatchGuard.com \
` bcollins@debian.org -- Ben.Collins@watchguard.com '
`---=========------=======-------------=-=-----=-===-======-------=--=---'
Re: Updated [patch] new part 1 for log command
Posted by Daniel Berlin <da...@dberlin.org>.
On Thu, 28 Feb 2002, Greg Stein wrote:
> [ per IRC, there are still some problems in the svn_client_log() handling
> for URLs; Ben is going to redo that portion ]
>
> The patch looks good, but for the URL code in svn_client_log. I'd recommend
> returning SVN_ERR_UNSUPPORTED_FEATURE for that for now, and we'll get the
> rest of this in. The URL can be done in a second round. Thoughts?
>
> On Thu, Feb 28, 2002 at 10:50:05PM -0500, Ben Collins wrote:
> > Here's a new patch. Changelog is the same. Fixes all suggestions, plus a
> > few.
>
> As I mentioned on IRC, when resubmitting a patch, please ensure to repost
> the log message. It is quite possible the person who will commit the patch
> has deleted the prior mail.
>
> >...
> > +++ ./subversion/include/svn_fs.h Thu Feb 28 21:52:50 2002
> > @@ -666,7 +666,9 @@
> >
> >
> > /* Allocate and return an array *REVS of svn_revnum_t revisions in
> > - which PATH under ROOT was modified. Use POOL for all allocations.
> > + which PATHS under ROOT were modified. Use POOL for all allocations.
> > + The array of *REVS are sorted in descending order. All duplicates
> > + will also be removed.
> >
> > NOTE: This function uses node-id ancestry alone to determine
> > modifiedness, and therefore does NOT claim that in any of the
> > @@ -674,7 +676,7 @@
> > directory entries lists changed, etc. */
> > svn_error_t *svn_fs_revisions_changed (apr_array_header_t **revs,
> > svn_fs_root_t *root,
> > - const char *path,
> > + const apr_array_header_t *paths,
> > apr_pool_t *pool);
>
> The doc string should tell the caller what the type of the PATHS elements
> are. Since that isn't type-checked by the compiler or anything, it is
> important to tell them. Just something like "the elements are 'const char *'"
>
> >...
> > +++ ./subversion/libsvn_fs/tree.c Thu Feb 28 22:17:05 2002
> >...
> > + {
> > + int t, exists = 0;
> > + svn_revnum_t revision;
> > + dag_node_t *node;
> > + int len = svn_fs_id_length (tmp_id);
> > +
> > + /* Get the dag node for this id. */
> > + SVN_ERR (svn_fs__dag_get_node (&node, args->fs, tmp_id, trail));
> > +
> > + /* Now get the revision from the dag. */
> > + SVN_ERR (svn_fs__dag_get_revision (&revision, node, trail));
> > +
> > + /* And put the revision into our array, if it doesn't already
> > + * exist. */
> > + for (t = 0; t < array->nelts && ! exists; t++)
> > + if (APR_ARRAY_IDX(array, t, svn_revnum_t) == revision)
> > + exists = 1;
>
> I'm mildly concerned about O(n^2) operation here, but I can't imagine this
> being all that large of an array. Regardless, when you find a match, you
> should 'break' out of the loop.
It does break , because !exists will not be true anymore.
> > + if (this_rev == ((svn_revnum_t *)(revs->elts))[i])
> > + matched = 1;
>
> And break out of here, too.
Same here, !matched will not be true anymore.
>
> Cheers,
> -g
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Updated [patch] new part 1 for log command
Posted by Greg Stein <gs...@lyra.org>.
[ per IRC, there are still some problems in the svn_client_log() handling
for URLs; Ben is going to redo that portion ]
The patch looks good, but for the URL code in svn_client_log. I'd recommend
returning SVN_ERR_UNSUPPORTED_FEATURE for that for now, and we'll get the
rest of this in. The URL can be done in a second round. Thoughts?
On Thu, Feb 28, 2002 at 10:50:05PM -0500, Ben Collins wrote:
> Here's a new patch. Changelog is the same. Fixes all suggestions, plus a
> few.
As I mentioned on IRC, when resubmitting a patch, please ensure to repost
the log message. It is quite possible the person who will commit the patch
has deleted the prior mail.
>...
> +++ ./subversion/include/svn_fs.h Thu Feb 28 21:52:50 2002
> @@ -666,7 +666,9 @@
>
>
> /* Allocate and return an array *REVS of svn_revnum_t revisions in
> - which PATH under ROOT was modified. Use POOL for all allocations.
> + which PATHS under ROOT were modified. Use POOL for all allocations.
> + The array of *REVS are sorted in descending order. All duplicates
> + will also be removed.
>
> NOTE: This function uses node-id ancestry alone to determine
> modifiedness, and therefore does NOT claim that in any of the
> @@ -674,7 +676,7 @@
> directory entries lists changed, etc. */
> svn_error_t *svn_fs_revisions_changed (apr_array_header_t **revs,
> svn_fs_root_t *root,
> - const char *path,
> + const apr_array_header_t *paths,
> apr_pool_t *pool);
The doc string should tell the caller what the type of the PATHS elements
are. Since that isn't type-checked by the compiler or anything, it is
important to tell them. Just something like "the elements are 'const char *'"
>...
> +++ ./subversion/libsvn_fs/tree.c Thu Feb 28 22:17:05 2002
>...
> + {
> + int t, exists = 0;
> + svn_revnum_t revision;
> + dag_node_t *node;
> + int len = svn_fs_id_length (tmp_id);
> +
> + /* Get the dag node for this id. */
> + SVN_ERR (svn_fs__dag_get_node (&node, args->fs, tmp_id, trail));
> +
> + /* Now get the revision from the dag. */
> + SVN_ERR (svn_fs__dag_get_revision (&revision, node, trail));
> +
> + /* And put the revision into our array, if it doesn't already
> + * exist. */
> + for (t = 0; t < array->nelts && ! exists; t++)
> + if (APR_ARRAY_IDX(array, t, svn_revnum_t) == revision)
> + exists = 1;
I'm mildly concerned about O(n^2) operation here, but I can't imagine this
being all that large of an array. Regardless, when you find a match, you
should 'break' out of the loop.
An alternative to the nested loop is to remove dups after the qsort()
occurs.
>...
> + /* We split the path here. If this is the only target (a full path)
> + we'll end up using BASENAME as the URL, and FILE as the path to
> + search on. For both cases, BASENAME gets passed to
> + svn_client__get_revision_number() below. */
> + svn_path_split (base, &basename, &file, pool);
>...
> @@ -94,13 +149,14 @@
> SVN_ERR (svn_client__open_ra_session (&session, ra_lib, URL, basename,
> TRUE, TRUE, auth_baton, pool));
>
> + /* Get the revisions based on the users "hints". */
> SVN_ERR (svn_client__get_revision_number
> (&start_revnum, ra_lib, session, start, basename->data, pool));
> SVN_ERR (svn_client__get_revision_number
> (&end_revnum, ra_lib, session, end, basename->data, pool));
When using a URL for an argument, there might not be a working copy. That
implies that NULL should be passed for the 'base_dir' arg to open_ra_session
(and FALSE for the do_store and use_admin flags). Also, NULL should be
passed for the 'path' argument to get_revision_number.
>...
> + if (revs)
> + {
> + int i, matched = 0;
> + for (i = 0; i < revs->nelts && !matched; i++)
> + if (this_rev == ((svn_revnum_t *)(revs->elts))[i])
> + matched = 1;
And break out of here, too.
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Updated [patch] new part 1 for log command
Posted by Ben Collins <bc...@debian.org>.
Here's a new patch. Changelog is the same. Fixes all suggestions, plus a
few.
--
.----------=======-=-======-=========-----------=====------------=-=-----.
/ Ben Collins -- Debian GNU/Linux -- WatchGuard.com \
` bcollins@debian.org -- Ben.Collins@watchguard.com '
`---=========------=======-------------=-=-----=-===-======-------=--=---'
Re: Updated [patch] new part 1 for log command
Posted by Greg Stein <gs...@lyra.org>.
On Thu, Feb 28, 2002 at 07:15:20PM -0500, Ben Collins wrote:
> On Thu, Feb 28, 2002 at 06:40:30PM -0500, Daniel Berlin wrote:
> >
> > On Thursday, February 28, 2002, at 06:10 PM, Ben Collins wrote:
> >
> > >This is an updated patch. Basically it expands log usage to allow for
> > >URL's and the -D option (which works with ra_local, but not ra_dav yet).
> > >Updated man page aswell.
> > >
> > > * libsvn_fs/tree.c(svn_fs_revisions_changed): Change to take an
> > > array of stringbuf's for paths instead of one path. It will then
> > > iterate to get a list of all revs for all paths (reduces amount
> > > of trails for getting info on multiple paths).
> > While you are in here, is there any reason to use stringbufs here (IE
> > could it be an array of const char *)
> > Do we ever modify the results?
>
> Can't do it. Some things do modify the paths. For instance, when taking
> off a component to get basename/dir, or adding components, or similar.
The FS API does not use stringbufs anywhere, except for 'svn_fs_unparse_id'.
I think it would be best if svn_fs_revisions_changed() took an array of
'const char *'. Especially given that all other repository paths are
modelled as that type. If you look at the code you added to
revisions_changed, you'll also note that you only use buf->data.
I would also point out that the svn_repos_get_logs() is not going to be
modifying the parameters, so those should (technically) be 'const char *'
also. However, that means some pushback against the callers which you
probably don't want to deal with at this stage.
Now, the practical result of making revisions_changed take 'const char *' is
that you'll need to map the get_logs' incoming array into a new array.
I'll review your patch now, but have these two points:
* the stringbuf in the FS API is effectively a showstopper
* you can simplify the apr_array_header_t access by using APR_ARRAY_IDX()
from the svn_types.h header.
More in a bit...
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Updated [patch] new part 1 for log command
Posted by Ben Collins <bc...@debian.org>.
On Thu, Feb 28, 2002 at 06:40:30PM -0500, Daniel Berlin wrote:
>
> On Thursday, February 28, 2002, at 06:10 PM, Ben Collins wrote:
>
> >This is an updated patch. Basically it expands log usage to allow for
> >URL's and the -D option (which works with ra_local, but not ra_dav yet).
> >Updated man page aswell.
> >
> > * libsvn_fs/tree.c(svn_fs_revisions_changed): Change to take an
> > array of stringbuf's for paths instead of one path. It will then
> > iterate to get a list of all revs for all paths (reduces amount
> > of trails for getting info on multiple paths).
> While you are in here, is there any reason to use stringbufs here (IE
> could it be an array of const char *)
> Do we ever modify the results?
Can't do it. Some things do modify the paths. For instance, when taking
off a component to get basename/dir, or adding components, or similar.
--
.----------=======-=-======-=========-----------=====------------=-=-----.
/ Ben Collins -- Debian GNU/Linux -- WatchGuard.com \
` bcollins@debian.org -- Ben.Collins@watchguard.com '
`---=========------=======-------------=-=-----=-===-======-------=--=---'
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Updated [patch] new part 1 for log command
Posted by Daniel Berlin <da...@dberlin.org>.
On Thursday, February 28, 2002, at 06:10 PM, Ben Collins wrote:
> This is an updated patch. Basically it expands log usage to allow for
> URL's and the -D option (which works with ra_local, but not ra_dav yet).
> Updated man page aswell.
>
> * libsvn_fs/tree.c(svn_fs_revisions_changed): Change to take an
> array of stringbuf's for paths instead of one path. It will then
> iterate to get a list of all revs for all paths (reduces amount
> of trails for getting info on multiple paths).
While you are in here, is there any reason to use stringbufs here (IE
could it be an array of const char *)
Do we ever modify the results?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org