Fix bug in combination of preferred and required content settings. When one was set to the empty string and the other set to some expression, this bug caused all files to be wanted, instead of only files matching the expression.

Avoid: MAny `MOr` otherexpression
Which matches anything.

5 files changed, 51 insertions, 2 deletions

diff --git a/Logs/PreferredContent.hs b/Logs/PreferredContent.hs
index a81c16f26..c21d67010 100644
--- a/Logs/PreferredContent.hs
+++ b/Logs/PreferredContent.hs
@@ -77,8 +77,9 @@ preferredRequiredMapsLoad = do
 		<$> Annex.Branch.get l
 	pc <- genmap preferredContentLog =<< groupPreferredContentMapRaw
 	rc <- genmap requiredContentLog M.empty
-	-- Required content is implicitly also preferred content, so OR
-	let m = M.unionWith MOr pc rc
+	-- Required content is implicitly also preferred content, so
+	-- combine.
+	let m = M.unionWith combineMatchers pc rc
 	Annex.changeState $ \s -> s
 		{ Annex.preferredcontentmap = Just m
 		, Annex.requiredcontentmap = Just rc
diff --git a/Utility/Matcher.hs b/Utility/Matcher.hs
index 19a77201c..d78e597dc 100644
--- a/Utility/Matcher.hs
+++ b/Utility/Matcher.hs
@@ -27,6 +27,7 @@ module Utility.Matcher (
 	matchM,
 	matchMrun,
 	isEmpty,
+	combineMatchers,
 
 	prop_matcher_sane
 ) where
@@ -142,6 +143,14 @@ isEmpty :: Matcher a -> Bool
 isEmpty MAny = True
 isEmpty _ = False
 
+{- Combines two matchers, yielding a matcher that will match anything
+ - both do. -}
+combineMatchers :: Matcher a -> Matcher a -> Matcher a
+combineMatchers a b 
+	| isEmpty a = b
+	| isEmpty b = a
+	| otherwise = a `MOr` b
+
 prop_matcher_sane :: Bool
 prop_matcher_sane = all (\m -> match dummy m ()) $ map generate
 	[ [Operation True]
diff --git a/debian/changelog b/debian/changelog
index f2f166ed7..74a73adb4 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -26,6 +26,10 @@ git-annex (5.20150825) UNRELEASED; urgency=medium
     in appropriate places.
   * Special remotes configured with autoenable=true will be automatically
     enabled when git-annex init is run.
+  * Fix bug in combination of preferred and required content settings.
+    When one was set to the empty string and the other set to some expression,
+    this bug caused all files to be wanted, instead of only files matching
+    the expression.
 
  -- Joey Hess <id@joeyh.name>  Tue, 01 Sep 2015 14:46:18 -0700
 
diff --git a/doc/bugs/Prefered_Content_not_Taken_into_Account.mdwn b/doc/bugs/Prefered_Content_not_Taken_into_Account.mdwn
index 818181dff..bdb3c9d97 100644
--- a/doc/bugs/Prefered_Content_not_Taken_into_Account.mdwn
+++ b/doc/bugs/Prefered_Content_not_Taken_into_Account.mdwn
@@ -360,3 +360,5 @@ On Ubuntu 14.04. All repos are on external USB drives on the same machine except
 
 
 [1] https://git-annex.branchable.com/forum/git-annex_does_not_respect_preferred_content_settings/
+
+> [[fixed|done]] --[[Joey]]
diff --git a/doc/bugs/Prefered_Content_not_Taken_into_Account/comment_1_70bccb282707b2e9b75c417832ed3459._comment b/doc/bugs/Prefered_Content_not_Taken_into_Account/comment_1_70bccb282707b2e9b75c417832ed3459._comment
new file mode 100644
index 000000000..ff8326d95
--- /dev/null
+++ b/doc/bugs/Prefered_Content_not_Taken_into_Account/comment_1_70bccb282707b2e9b75c417832ed3459._comment
@@ -0,0 +1,33 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 1"""
+ date="2015-09-15T16:04:28Z"
+ content="""
+I find it's not useful to split a conversation between multiple different
+pages, so this bug is a bit counterproductive. 
+
+But, you made up for that by posting the whole vicfg, which I was somehow
+able to notice, amoungst all the noise, has this unusual bit:
+
+	required 132503d4-dcde-4790-aabb-ee5ba539a3a0 =
+	# (for ozge)
+	required 1e1d0c4e-b1da-465f-9140-7128a7e3ee13 =
+	# (for irem)
+	required aabc3536-a423-42b6-a234-5f110607296e =
+	# (for yesim)
+	required ba3593c0-ddf1-4433-9916-aa25d1a52895 =
+	# (for buse [origin])
+	required bff7238e-bd92-4929-88a8-c59c1a1dcf03 =
+
+So, required content has been set to "". It turns out that when this is done,
+git-annex thinks that all files are preferred! This is because of a bug
+when combining the required content and preferred content expressions.
+
+I reproduced this; `git annex get --auto` was not getting a file,
+which was already known to be in another PodA repository. Then I ran `git annex required . ""`
+and `git annex get --auto` started getting all files.
+
+You can't unset a required content setting back to being commented out.
+A reasonable workaround is to set those to "groupwanted" too. Or get the
+next git-annex release.
+"""]]