more gracefully handle overlapping ignore options

Summary:
facebook/buck#170 (comment) refers
to issues watching .svn dirs and requests a way to ignore those.

facebook/buck#170 (comment)
demonstrates that someone got bitten by queries stopping working when
they were experimenting with ignores.

While I don't think ignoring .git or .hg makes a big difference in the
grand scheme (the default ignore_vcs works very well on our large repos,
and noticing VCS activity is an important part of filesystem settle
detection for triggers), it sounds like doing so for .svn is more of an
issue.

What this diff does:

if a dir is listed in both `ignore_dirs` and `ignore_vcs`, we'll take
the `ignore_dirs` placement as definitive, and elide that dir from the
effective value of `ignore_vcs`.

This means that it will no longer be candidate location for the cookie
dir and that queries will continue to work just fine.

Addresses #17

Test Plan: `make integration`

Reviewers: fsilva, jimp, sid0

Reviewed By: sid0

Differential Revision: https://phabricator.fb.com/D1484322

docs/configuration.markdown

@@ -127,6 +127,12 @@ would ignore the `build` directory at the top level of the watched tree, and
 everything below it.  It will never appear in the watchman query results for
 the tree.
 
+Since version 2.9.9, if you list a dir in `ignore_dirs` that is also listed in
+`ignore_vcs`, the `ignore_dirs` placement will take precedence.  This may not
+sound like a big deal, but since `ignore_vcs` is used as a hint to for the
+placement of [cookie files](/watchman/docs/cookies.html), having these two
+options overlap in earlier versions would break watchman queries.
+
 #### gc_age_seconds
 
 Deleted files (and dirs) older than this are periodically pruned from the

root.c

@@ -251,6 +251,15 @@ static bool apply_ignore_vcs_configuration(w_root_t *root, char **errmsg)
 
     name = w_string_new(ignore);
     fullname = w_string_path_cat(root->root_path, name);
+
+    // if we are completely ignoring this dir, we have nothing more to
+    // do here
+    if (w_ht_get(root->ignore_dirs, w_ht_ptr_val(fullname))) {
+      w_string_delref(fullname);
+      w_string_delref(name);
+      continue;
+    }
+
     w_ht_set(root->ignore_vcs, w_ht_ptr_val(fullname),
         w_ht_ptr_val(fullname));
 
@@ -360,13 +369,13 @@ static w_root_t *w_root_new(const char *path, char **errmsg)
   root->gc_interval = cfg_get_int(root, "gc_interval_seconds",
       DEFAULT_GC_INTERVAL);
 
+  apply_ignore_configuration(root);
+
   if (!apply_ignore_vcs_configuration(root, errmsg)) {
     w_root_delref(root);
     return NULL;
   }
 
-  apply_ignore_configuration(root);
-
   if (!w_root_init(root, errmsg)) {
     w_root_delref(root);
     return NULL;

tests/integration/ignore.php

@@ -48,6 +48,22 @@ function testInvalidIgnore() {
     }
   }
 
+  function testIgnoreOverlapVCSIgnore() {
+    $dir = PhutilDirectoryFixture::newEmptyFixture();
+    $root = realpath($dir->getPath());
+
+    $cfg = array(
+      'ignore_dirs' => array('.hg'),
+    );
+    file_put_contents("$root/.watchmanconfig", json_encode($cfg));
+    mkdir("$root/.hg");
+
+    $this->watch($root);
+    $this->assertFileList($root, array('.watchmanconfig'));
+    touch("$root/foo");
+    $this->assertFileList($root, array('.watchmanconfig', 'foo'));
+  }
+
   function testIgnoreGeneric() {
     $dir = PhutilDirectoryFixture::newEmptyFixture();
     $root = realpath($dir->getPath());