X-Git-Url: https://git.madduck.net/code/myrepos.git/blobdiff_plain/cab31c75dc608d3567c07c616c2cd84f7be2ce3e..59439878f5a85ff5ab6846106079549f3325d8ae:/TODO

diff --git a/TODO b/TODO
index 000d5ab..86eaf5a 100644
--- a/TODO
+++ b/TODO
@@ -1,29 +1,27 @@
 * more revision control systems
 
-* support for tracking repo renames
+* a way to detect repos in a tree that are not registered, and warn
+  about or even auto-register them. (svn externals make this quite
+  difficult!)
 
-  It should be possible to tell mr that there used to be a repo at
-  src/foo/bar, and it's been moved to src/bar. mr would then detect if the
-  move needs to be done, and handle it. This is mostly useful when mrconfig
-  files are shared accross several systems.
+* When there are chained mrconfig files, mr could be smarter about
+  checkouts and updates. Ie, when a new version of an mrconfig file is
+  checked out or updated, throw all the info from the old one away, and
+  process the new one.
 
-  [src/bar]
-  renamedfrom = src/foo/bar
+  Until this is fixed, checkouts and updates need to be manually repeated
+  after mrconfig files have changes.
 
-  (Support multple renames of a single repo?)
+* offline support
 
-* mr register
+  If I commit something to git while offline, it would be nice if mr could
+  have a way to push that change when I get online.
 
-  Idea is you check out a repo and then use mr register to add it to the
-  closest mrconfig file.
+  One approach would be to notice when mr commit fails, and queue the
+  commit up to be tried happen again when "mr retry" is run. This could
+  also notice other failing commands, such as "mr up".
 
-  mr register would be implemented as a shell command that then calls
-  mr config with flags that make it edit the mrconfig file:
-
-  if [ -d "$MR_REPO/.svn" ]; then
-  	url=$(svn info "$MR_REPO" | grep -i ^URL: | cut -d ' ' -f 2)
-	if [ -z "$url" ]; then
-		error "cannot determine svn url"
-	fi
-	mr -c "$MR_CONFIG" config --add "$MR_REPO" --checkout="svn co $URL"
-  fi
+  Would it make sense to have to first run "mr offline", before mr starts
+  recording such failures? If so, "mr online" would be the thing to run
+  when getting back online, this would both retry queued commands, and stop
+  queuing new failures.