git-p4: fixing --changes-block-size handlingld/p4-changes-block-size

The --changes-block-size handling was intended to help when
a user has a limited "maxscanrows" (see "p4 group"). It used
"p4 changes -m $maxchanges" to limit the number of results.

Unfortunately, it turns out that the "maxscanrows" and "maxresults"
limits are actually applied *before* the "-m maxchanges" parameter
is considered (experimentally).

Fix the block-size handling so that it gets blocks of changes
limited by revision number ($Start..$Start+$N, etc). This limits
the number of results early enough that both sets of tests pass.

Note that many other Perforce operations can fail for the same
reason (p4 print, p4 files, etc) and it's probably not possible
to workaround this. In the real world, this is probably not
usually a problem.

Signed-off-by: Luke Diamand <luke@diamand.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

1 files changed, 63 insertions, 22 deletions

diff --git a/git-p4.py b/git-p4.py
index ca6bb95c57..73f05f0a88 100755
--- a/git-p4.py
+++ b/git-p4.py
@@ -43,6 +43,9 @@ verbose = False
 # Only labels/tags matching this will be imported/exported
 defaultLabelRegexp = r'[a-zA-Z0-9_\-.]+$'
 
+# Grab changes in blocks of this many revisions, unless otherwise requested
+defaultBlockSize = 512
+
 def p4_build_cmd(cmd):
     """Build a suitable p4 command line.
 
@@ -249,6 +252,10 @@ def p4_reopen(type, f):
 def p4_move(src, dest):
     p4_system(["move", "-k", wildcard_encode(src), wildcard_encode(dest)])
 
+def p4_last_change():
+    results = p4CmdList(["changes", "-m", "1"])
+    return int(results[0]['change'])
+
 def p4_describe(change):
     """Make sure it returns a valid result by checking for
        the presence of field "time".  Return a dict of the
@@ -742,43 +749,77 @@ def createOrUpdateBranchesFromOrigin(localRefPrefix = "refs/remotes/p4/", silent
 def originP4BranchesExist():
         return gitBranchExists("origin") or gitBranchExists("origin/p4") or gitBranchExists("origin/p4/master")
 
-def p4ChangesForPaths(depotPaths, changeRange, block_size):
+
+def p4ParseNumericChangeRange(parts):
+    changeStart = int(parts[0][1:])
+    if parts[1] == '#head':
+        changeEnd = p4_last_change()
+    else:
+        changeEnd = int(parts[1])
+
+    return (changeStart, changeEnd)
+
+def chooseBlockSize(blockSize):
+    if blockSize:
+        return blockSize
+    else:
+        return defaultBlockSize
+
+def p4ChangesForPaths(depotPaths, changeRange, requestedBlockSize):
     assert depotPaths
-    assert block_size
 
-    # Parse the change range into start and end
+    # Parse the change range into start and end. Try to find integer
+    # revision ranges as these can be broken up into blocks to avoid
+    # hitting server-side limits (maxrows, maxscanresults). But if
+    # that doesn't work, fall back to using the raw revision specifier
+    # strings, without using block mode.
+
     if changeRange is None or changeRange == '':
-        changeStart = '@1'
-        changeEnd = '#head'
+        changeStart = 1
+        changeEnd = p4_last_change()
+        block_size = chooseBlockSize(requestedBlockSize)
     else:
         parts = changeRange.split(',')
         assert len(parts) == 2
-        changeStart = parts[0]
-        changeEnd = parts[1]
+        try:
+            (changeStart, changeEnd) = p4ParseNumericChangeRange(parts)
+            block_size = chooseBlockSize(requestedBlockSize)
+        except:
+            changeStart = parts[0][1:]
+            changeEnd = parts[1]
+            if requestedBlockSize:
+                die("cannot use --changes-block-size with non-numeric revisions")
+            block_size = None
 
     # Accumulate change numbers in a dictionary to avoid duplicates
     changes = {}
 
     for p in depotPaths:
         # Retrieve changes a block at a time, to prevent running
-        # into a MaxScanRows error from the server.
-        start = changeStart
-        end = changeEnd
-        get_another_block = True
-        while get_another_block:
-            new_changes = []
+        # into a MaxResults/MaxScanRows error from the server.
+
+        while True:
             cmd = ['changes']
-            cmd += ['-m', str(block_size)]
-            cmd += ["%s...%s,%s" % (p, start, end)]
+
+            if block_size:
+                end = min(changeEnd, changeStart + block_size)
+                revisionRange = "%d,%d" % (changeStart, end)
+            else:
+                revisionRange = "%s,%s" % (changeStart, changeEnd)
+
+            cmd += ["%s...@%s" % (p, revisionRange)]
+
             for line in p4_read_pipe_lines(cmd):
                 changeNum = int(line.split(" ")[1])
-                new_changes.append(changeNum)
                 changes[changeNum] = True
-            if len(new_changes) == block_size:
-                get_another_block = True
-                end = '@' + str(min(new_changes))
-            else:
-                get_another_block = False
+
+            if not block_size:
+                break
+
+            if end >= changeEnd:
+                break
+
+            changeStart = end + 1
 
     changelist = changes.keys()
     changelist.sort()
@@ -1974,7 +2015,7 @@ class P4Sync(Command, P4UserMap):
         self.syncWithOrigin = True
         self.importIntoRemotes = True
         self.maxChanges = ""
-        self.changes_block_size = 500
+        self.changes_block_size = None
         self.keepRepoPath = False
         self.depotPaths = None
         self.p4BranchesInGit = []