pathauditor: disable cache of audited paths by default (issue5628)

The initial attempt was to discard cache when appropriate, but it appears
to be error prone. We had to carefully inspect all places where audit() is
called e.g. without actually updating filesystem, before removing files and
directories, etc.

So, this patch disables the cache of audited paths by default, and enables
it only for the following cases:

 - short-lived auditor objects
 - repo.vfs, repo.svfs, and repo.cachevfs, which are managed directories
   and considered sort of append-only (a file/directory would never be
   replaced with a symlink)

There would be more cacheable vfs objects (e.g. mq.queue.opener), but I
decided not to inspect all of them in this patch. We can make them cached
later.

Benchmark result:

- using old clone of http://selenic.com/repo/linux-2.6/ (38319 files)
- on tmpfs
- run HGRCPATH=/dev/null hg up -q --time tip && hg up -q null
- try 4 times and take the last three results

original:
real 7.480 secs (user 1.140+22.760 sys 0.150+1.690)
real 8.010 secs (user 1.070+22.280 sys 0.170+2.120)
real 7.470 secs (user 1.120+22.390 sys 0.120+1.910)

clearcache (the other series):
real 7.680 secs (user 1.120+23.420 sys 0.140+1.970)
real 7.670 secs (user 1.110+23.620 sys 0.130+1.810)
real 7.740 secs (user 1.090+23.510 sys 0.160+1.940)

enable cache only for vfs and svfs (this series):
real 8.730 secs (user 1.500+25.190 sys 0.260+2.260)
real 8.750 secs (user 1.490+25.170 sys 0.250+2.340)
real 9.010 secs (user 1.680+25.340 sys 0.280+2.540)

remove cache function at all (for reference):
real 9.620 secs (user 1.440+27.120 sys 0.250+2.980)
real 9.420 secs (user 1.400+26.940 sys 0.320+3.130)
real 9.760 secs (user 1.530+27.270 sys 0.250+2.970)

hgext/fsmonitor/__init__.py

@@ -394,7 +394,7 b' def overridewalk(orig, self, match, subr'
             visit.update(f for f in copymap
                          if f not in results and matchfn(f))
 
-    audit = pathutil.pathauditor(self._root).check
+    audit = pathutil.pathauditor(self._root, cached=True).check
     auditpass = [f for f in visit if audit(f)]
     auditpass.sort()
     auditfail = visit.difference(auditpass)

mercurial/cmdutil.py

@@ -3218,7 +3218,7 b' def _performrevert(repo, parents, ctx, a'
             pass
         repo.dirstate.remove(f)
 
-    audit_path = pathutil.pathauditor(repo.root)
+    audit_path = pathutil.pathauditor(repo.root, cached=True)
     for f in actions['forget'][0]:
         if interactive:
             choice = repo.ui.promptchoice(

mercurial/dirstate.py

@@ -1089,7 +1089,7 b' class dirstate(object):'
                 # that wasn't ignored, and everything that matched was stat'ed
                 # and is already in results.
                 # The rest must thus be ignored or under a symlink.
-                audit_path = pathutil.pathauditor(self._root)
+                audit_path = pathutil.pathauditor(self._root, cached=True)
 
                 for nf in iter(visit):
                     # If a stat for the same file was already added with a

mercurial/localrepo.py

@@ -273,8 +273,8 b' class localrepository(object):'
         self.origroot = path
         self.auditor = pathutil.pathauditor(self.root, self._checknested)
         self.nofsauditor = pathutil.pathauditor(self.root, self._checknested,
-                                                realfs=False)
-        self.vfs = vfsmod.vfs(self.path)
+                                                realfs=False, cached=True)
+        self.vfs = vfsmod.vfs(self.path, cacheaudited=True)
         self.baseui = baseui
         self.ui = baseui.copy()
         self.ui.copy = baseui.copy # prevent copying repo configuration
@@ -350,7 +350,8 b' class localrepository(object):'
                 raise
 
         self.store = store.store(
-                self.requirements, self.sharedpath, vfsmod.vfs)
+            self.requirements, self.sharedpath,
+            lambda base: vfsmod.vfs(base, cacheaudited=True))
         self.spath = self.store.path
         self.svfs = self.store.vfs
         self.sjoin = self.store.join

mercurial/pathutil.py

@@ -33,13 +33,18 b' class pathauditor(object):'
     The file system checks are only done when 'realfs' is set to True (the
     default). They should be disable then we are auditing path for operation on
     stored history.
+
+    If 'cached' is set to True, audited paths and sub-directories are cached.
+    Be careful to not keep the cache of unmanaged directories for long because
+    audited paths may be replaced with symlinks.
     '''
 
-    def __init__(self, root, callback=None, realfs=True):
+    def __init__(self, root, callback=None, realfs=True, cached=False):
         self.audited = set()
         self.auditeddir = set()
         self.root = root
         self._realfs = realfs
+        self._cached = cached
         self.callback = callback
         if os.path.lexists(root) and not util.fscasesensitive(root):
             self.normcase = util.normcase
@@ -96,10 +101,11 b' class pathauditor(object):'
                 self._checkfs(prefix, path)
             prefixes.append(normprefix)
 
-        self.audited.add(normpath)
-        # only add prefixes to the cache after checking everything: we don't
-        # want to add "foo/bar/baz" before checking if there's a "foo/.hg"
-        self.auditeddir.update(prefixes)
+        if self._cached:
+            self.audited.add(normpath)
+            # only add prefixes to the cache after checking everything: we don't
+            # want to add "foo/bar/baz" before checking if there's a "foo/.hg"
+            self.auditeddir.update(prefixes)
 
     def _checkfs(self, prefix, path):
         """raise exception if a file system backed check fails"""

mercurial/scmutil.py

@@ -667,7 +667,7 b' def _interestingfiles(repo, matcher):'
     This is different from dirstate.status because it doesn't care about
     whether files are modified or clean.'''
     added, unknown, deleted, removed, forgotten = [], [], [], [], []
-    audit_path = pathutil.pathauditor(repo.root)
+    audit_path = pathutil.pathauditor(repo.root, cached=True)
 
     ctx = repo[None]
     dirstate = repo.dirstate

mercurial/vfs.py

@@ -276,13 +276,19 b' class vfs(abstractvfs):'
 
     This class is used to hide the details of COW semantics and
     remote file access from higher level code.
+
+    'cacheaudited' should be enabled only if (a) vfs object is short-lived, or
+    (b) the base directory is managed by hg and considered sort-of append-only.
+    See pathutil.pathauditor() for details.
     '''
-    def __init__(self, base, audit=True, expandpath=False, realpath=False):
+    def __init__(self, base, audit=True, cacheaudited=False, expandpath=False,
+                 realpath=False):
         if expandpath:
             base = util.expandpath(base)
         if realpath:
             base = os.path.realpath(base)
         self.base = base
+        self._cacheaudited = cacheaudited
         self.mustaudit = audit
         self.createmode = None
         self._trustnlink = None
@@ -295,7 +301,8 b' class vfs(abstractvfs):'
     def mustaudit(self, onoff):
         self._audit = onoff
         if onoff:
-            self.audit = pathutil.pathauditor(self.base)
+            self.audit = pathutil.pathauditor(
+                self.base, cached=self._cacheaudited)
         else:
             self.audit = util.always
 

tests/test-audit-path.t

@@ -169,9 +169,9 b' and the rebase should fail (issue5628)'
   $ hg up -qC 2
   $ hg rebase -s 2 -d 1 --config extensions.rebase=
   rebasing 2:e73c21d6b244 "file a/poisoned" (tip)
-  saved backup bundle to * (glob)
+  abort: path 'a/poisoned' traverses symbolic link 'a'
+  [255]
   $ ls ../merge-symlink-out
-  poisoned
 
   $ cd ..
 
@@ -211,10 +211,9 b' audited first by calculateupdates(), whe'
 
   $ hg up -qC null
   $ hg up 1
-  2 files updated, 0 files merged, 0 files removed, 0 files unresolved
+  abort: path 'a/b' traverses symbolic link 'a'
+  [255]
   $ ls ../update-symlink-out
-  b
-  $ rm ../update-symlink-out/b
 
 try branch update replacing directory with symlink, and its content: the
 path 'a' is audited as a directory first, which should be audited again as
@@ -223,9 +222,9 b' a symlink.'
   $ rm -f a
   $ hg up -qC 2
   $ hg up 1
-  2 files updated, 0 files merged, 1 files removed, 0 files unresolved
+  abort: path 'a/b' traverses symbolic link 'a'
+  [255]
   $ ls ../update-symlink-out
-  b
 
   $ cd ..
 

tests/test-commandserver.t

@@ -956,10 +956,9 b' and the merge should fail (issue5628)'
   *** runcommand up -qC 2
   *** runcommand up -qC 1
   *** runcommand merge 2
-  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
-  (branch merge, don't forget to commit)
+  abort: path 'a/poisoned' traverses symbolic link 'a'
+   [255]
   $ ls ../merge-symlink-out
-  poisoned
 
 cache of repo.auditor should be discarded, so matcher would never traverse
 symlinks:
@@ -983,7 +982,8 b' symlinks:'
   *** runcommand up -qC 0
   *** runcommand up -qC 1
   *** runcommand files a/poisoned
-   [1]
+  abort: path 'a/poisoned' traverses symbolic link 'a'
+   [255]
 
   $ cd ..