fix parsing of tags. make parse errors useful. add new tag tests.

old code read every head of .hgtags. delete and recreate of .hgtags gave
new head, but if error in deleted rev, .hgtags had error messages every
time it was parsed. this was very hard to fix, because deleted revs hard
to get back and update, needed merges too.

new code reads .hgtags on every head. advantage is if parse error
happens with new code, is possible to fix them by editing .hgtags on a
head and committing.

NOTE: new code uses binary search of manifest of each head to be fast,
but still much slower than old code. best thing would be to have delete
record stored in filelog so we never touch manifest. could find live
heads directly from filelog. this is more work than i want now.

new tests check for parse of tags on different heads, and inaccessible
heads created by delete and recreate of .hgtags.

mercurial/localrepo.py

@@ -166,37 +166,44 b' class localrepository(object):'
                     return
                 s = l.split(" ", 1)
                 if len(s) != 2:
-                    self.ui.warn(_("%s: ignoring invalid tag\n") % context)
+                    self.ui.warn(_("%s: cannot parse entry\n") % context)
                     return
                 node, key = s
+                key = key.strip()
                 try:
                     bin_n = bin(node)
                 except TypeError:
-                    self.ui.warn(_("%s: ignoring invalid tag\n") % context)
+                    self.ui.warn(_("%s: node '%s' is not well formed\n") %
+                                 (context, node))
                     return
                 if bin_n not in self.changelog.nodemap:
-                    self.ui.warn(_("%s: ignoring invalid tag\n") % context)
+                    self.ui.warn(_("%s: tag '%s' refers to unknown node\n") %
+                                 (context, key))
                     return
-                self.tagscache[key.strip()] = bin_n
+                self.tagscache[key] = bin_n
 
-            # read each head of the tags file, ending with the tip
+            # read the tags file from each head, ending with the tip,
             # and add each tag found to the map, with "newer" ones
             # taking precedence
+            heads = self.heads()
+            heads.reverse()
             fl = self.file(".hgtags")
-            h = fl.heads()
-            h.reverse()
-            for r in h:
+            for node in heads:
+                change = self.changelog.read(node)
+                rev = self.changelog.rev(node)
+                fn, ff = self.manifest.find(change[0], '.hgtags')
+                if fn is None: continue
                 count = 0
-                for l in fl.read(r).splitlines():
+                for l in fl.read(fn).splitlines():
                     count += 1
-                    parsetag(l, ".hgtags:%d" % count)
-
+                    parsetag(l, _(".hgtags (rev %d:%s), line %d") %
+                             (rev, short(node), count))
             try:
                 f = self.opener("localtags")
                 count = 0
                 for l in f:
                     count += 1
-                    parsetag(l, "localtags:%d" % count)
+                    parsetag(l, _("localtags, line %d") % count)
             except IOError:
                 pass
 

mercurial/manifest.py

@@ -43,48 +43,61 b' class manifest(revlog):'
     def diff(self, a, b):
         return mdiff.textdiff(str(a), str(b))
 
+    def _search(self, m, s, lo=0, hi=None):
+        '''return a tuple (start, end) that says where to find s within m.
+
+        If the string is found m[start:end] are the line containing
+        that string.  If start == end the string was not found and
+        they indicate the proper sorted insertion point.  This was
+        taken from bisect_left, and modified to find line start/end as
+        it goes along.
+
+        m should be a buffer or a string
+        s is a string'''
+        def advance(i, c):
+            while i < lenm and m[i] != c:
+                i += 1
+            return i
+        lenm = len(m)
+        if not hi:
+            hi = lenm
+        while lo < hi:
+            mid = (lo + hi) // 2
+            start = mid
+            while start > 0 and m[start-1] != '\n':
+                start -= 1
+            end = advance(start, '\0')
+            if m[start:end] < s:
+                # we know that after the null there are 40 bytes of sha1
+                # this translates to the bisect lo = mid + 1
+                lo = advance(end + 40, '\n') + 1
+            else:
+                # this translates to the bisect hi = mid
+                hi = start
+        end = advance(lo, '\0')
+        found = m[lo:end]
+        if cmp(s, found) == 0:
+            # we know that after the null there are 40 bytes of sha1
+            end = advance(end + 40, '\n')
+            return (lo, end+1)
+        else:
+            return (lo, lo)
+
+    def find(self, node, f):
+        '''look up entry for a single file efficiently.
+        return (node, flag) pair if found, (None, None) if not.'''
+        if self.mapcache and node == self.mapcache[0]:
+            return self.mapcache[1].get(f), self.mapcache[2].get(f)
+        text = self.revision(node)
+        start, end = self._search(text, f)
+        if start == end:
+            return None, None
+        l = text[start:end]
+        f, n = l.split('\0')
+        return bin(n[:40]), n[40:-1] == 'x'
+
     def add(self, map, flags, transaction, link, p1=None, p2=None,
             changed=None):
-
-        # returns a tuple (start, end).  If the string is found
-        # m[start:end] are the line containing that string.  If start == end
-        # the string was not found and they indicate the proper sorted 
-        # insertion point.  This was taken from bisect_left, and modified
-        # to find line start/end as it goes along.
-        #
-        # m should be a buffer or a string
-        # s is a string
-        #
-        def manifestsearch(m, s, lo=0, hi=None):
-            def advance(i, c):
-                while i < lenm and m[i] != c:
-                    i += 1
-                return i
-            lenm = len(m)
-            if not hi:
-                hi = lenm
-            while lo < hi:
-                mid = (lo + hi) // 2
-                start = mid
-                while start > 0 and m[start-1] != '\n':
-                    start -= 1
-                end = advance(start, '\0')
-                if m[start:end] < s:
-                    # we know that after the null there are 40 bytes of sha1
-                    # this translates to the bisect lo = mid + 1
-                    lo = advance(end + 40, '\n') + 1
-                else:
-                    # this translates to the bisect hi = mid
-                    hi = start
-            end = advance(lo, '\0')
-            found = m[lo:end]
-            if cmp(s, found) == 0: 
-                # we know that after the null there are 40 bytes of sha1
-                end = advance(end + 40, '\n')
-                return (lo, end+1)
-            else:
-                return (lo, lo)
-
         # apply the changes collected during the bisect loop to our addlist
         # return a delta suitable for addrevision
         def addlistdelta(addlist, x):
@@ -137,7 +150,7 b' class manifest(revlog):'
             for w in work:
                 f = w[0]
                 # bs will either be the index of the item or the insert point
-                start, end = manifestsearch(addbuf, f, start)
+                start, end = self._search(addbuf, f, start)
                 if w[1] == 0:
                     l = "%s\000%s%s\n" % (f, hex(map[f]),
                                           flags[f] and "x" or '')

tests/test-tags

@@ -32,12 +32,31 b' hg id'
 hg status
 
 hg commit -m "merge" -d "1000000 0"
+
+# create fake head, make sure tag not visible afterwards
+cp .hgtags tags
+hg tag -d "1000000 0" last
+hg rm .hgtags
+hg commit -m "remove" -d "1000000 0"
+
+mv tags .hgtags
+hg add .hgtags
+hg commit -m "readd" -d "1000000 0"
+
+hg tags
+
 # invalid tags
 echo "spam" >> .hgtags
 echo >> .hgtags
 echo "foo bar" >> .hgtags
 echo "$T invalid" | sed "s/..../a5a5/" >> .hg/localtags
 hg commit -m "tags" -d "1000000 0"
+
+# report tag parse error on other head
+hg up 3
+echo 'x y' >> .hgtags
+hg commit -m "head" -d "1000000 0"
+
 hg tags
 hg tip
 

tests/test-tags.out

@@ -16,17 +16,26 b' 1 files updated, 0 files merged, 0 files'
 (branch merge, don't forget to commit)
 8216907a933d+8a3ca90d111d+ tip
 M .hgtags
-.hgtags:2: ignoring invalid tag
-.hgtags:4: ignoring invalid tag
-localtags:1: ignoring invalid tag
-tip                                4:fd868a874787a7b5af31e1675666ce691c803035
+tip                                6:c6af9d771a81bb9c7f267ec03491224a9f8ba1cd
 first                              0:0acdaf8983679e0aac16e811534eb49d7ee1f2b4
-changeset:   4:fd868a874787
-.hgtags:2: ignoring invalid tag
-.hgtags:4: ignoring invalid tag
-localtags:1: ignoring invalid tag
+.hgtags (rev 7:39bba1bbbc4c), line 2: cannot parse entry
+.hgtags (rev 7:39bba1bbbc4c), line 4: node 'foo' is not well formed
+localtags, line 1: tag 'invalid' refers to unknown node
+1 files updated, 0 files merged, 0 files removed, 0 files unresolved
+.hgtags (rev 7:39bba1bbbc4c), line 2: cannot parse entry
+.hgtags (rev 7:39bba1bbbc4c), line 4: node 'foo' is not well formed
+.hgtags (rev 8:4ca6f1b1a68c), line 2: node 'x' is not well formed
+localtags, line 1: tag 'invalid' refers to unknown node
+tip                                8:4ca6f1b1a68c77be687a03aaeb1614671ba59b20
+first                              0:0acdaf8983679e0aac16e811534eb49d7ee1f2b4
+changeset:   8:4ca6f1b1a68c
+.hgtags (rev 7:39bba1bbbc4c), line 2: cannot parse entry
+.hgtags (rev 7:39bba1bbbc4c), line 4: node 'foo' is not well formed
+.hgtags (rev 8:4ca6f1b1a68c), line 2: node 'x' is not well formed
+localtags, line 1: tag 'invalid' refers to unknown node
 tag:         tip
+parent:      3:b2ef3841386b
 user:        test
 date:        Mon Jan 12 13:46:40 1970 +0000
-summary:     tags
+summary:     head