convert: ignore hg source errors with hg.ignoreerrors (issue 1357)

This flag makes missing revlog errors to be ignored which allow broken
repositories to be fixed by converting them from and to Mercurial.

hgext/convert/__init__.py

@@ -85,6 +85,10 b' def convert(ui, src, dest=None, revmapfi'
     Mercurial Source
     -----------------
 
+    --config convert.hg.ignoreerrors=False (boolean) 
+        ignore integrity errors when reading. Use it to fix Mercurial
+        repositories with missing revlogs, by converting from and to
+        Mercurial.
     --config convert.hg.saverev=True          (boolean)
         allow target to preserve source revision ID
     --config convert.hg.startrev=0            (hg revision identifier)

hgext/convert/hg.py

@@ -192,6 +192,8 b' class mercurial_sink(converter_sink):'
 class mercurial_source(converter_source):
     def __init__(self, ui, path, rev=None):
         converter_source.__init__(self, ui, path, rev)
+        self.ignoreerrors = ui.configbool('convert', 'hg.ignoreerrors', False)
+        self.ignored = {}
         self.saverev = ui.configbool('convert', 'hg.saverev', True)
         try:
             self.repo = hg.repository(self.ui, path)
@@ -253,23 +255,35 b' class mercurial_source(converter_source)'
         parents = self.parents(ctx)
         if not parents:
             files = util.sort(ctx.manifest().keys())
-            return [(f, rev) for f in files], {}
+            return [(f, rev) for f in files if f not in self.ignored], {}
         if self._changescache and self._changescache[0] == rev:
             m, a, r = self._changescache[1]
         else:
             m, a, r = self.repo.status(parents[0], ctx.node())[:3]
-        changes = [(name, rev) for name in m + a + r]
-        return util.sort(changes), self.getcopies(ctx, m + a)
+        # getcopies() detects missing revlogs early, run it before
+        # filtering the changes.
+        copies = self.getcopies(ctx, m + a)
+        changes = [(name, rev) for name in m + a + r 
+                   if name not in self.ignored]
+        return util.sort(changes), copies
 
     def getcopies(self, ctx, files):
         copies = {}
         for name in files:
+            if name in self.ignored:
+                continue
             try:
-                copynode = ctx.filectx(name).renamed()[0]
-                if self.keep(copynode):
-                    copies[name] = copynode
+                copysource, copynode = ctx.filectx(name).renamed()
+                if copysource in self.ignored or not self.keep(copynode):
+                    continue
+                copies[name] = copysource
             except TypeError:
                 pass
+            except revlog.LookupError, e:
+                if not self.ignoreerrors:
+                    raise
+                self.ignored[name] = 1
+                self.ui.warn(_('ignoring: %s\n') % e)
         return copies
 
     def getcommit(self, rev):
@@ -297,6 +311,7 b' class mercurial_source(converter_source)'
         else:
             i = i or 0
             changes = self.repo.status(parents[i], ctx.node())[:3]
+        changes = [[f for f in l if f not in self.ignored] for l in changes]
 
         if i == 0:
             self._changescache = (rev, changes)

tests/test-convert-hg-source

@@ -36,5 +36,34 b' cd ..'
 hg convert --datesort orig new 2>&1 | grep -v 'subversion python bindings could not be loaded'
 cd new
 hg out ../orig
+cd ..
+
+echo % init broken repository
+hg init broken
+cd broken
+echo a >> a
+echo b >> b
+hg ci -qAm init
+echo a >> a
+echo b >> b
+hg copy b c
+hg ci -qAm changeall
+hg up -qC 0
+echo bc >> b
+hg ci -m changebagain
+HGMERGE=internal:local hg -q merge
+hg ci -m merge
+hg mv b d
+hg ci -m moveb
+echo % break it
+rm .hg/store/data/b.*
+cd ..
+
+hg --config convert.hg.ignoreerrors=True convert broken fixed
+hg -R fixed verify
+echo '% manifest -r 0'
+hg -R fixed manifest -r 0
+echo '% manifest -r tip'
+hg -R fixed manifest -r tip
 
 true

tests/test-convert-hg-source.out

@@ -20,3 +20,27 b' 0 mark baz executable'
 comparing with ../orig
 searching for changes
 no changes found
+% init broken repository
+created new head
+% break it
+initializing destination fixed repository
+scanning source...
+sorting...
+converting...
+4 init
+3 changebagain
+ignoring: data/b.i@4b3c32ced4f8: no match found
+2 changeall
+1 merge
+0 moveb
+checking changesets
+checking manifests
+crosschecking files in changesets and manifests
+checking files
+3 files, 5 changesets, 5 total revisions
+% manifest -r 0
+a
+% manifest -r tip
+a
+c
+d

tests/test-convert.out

@@ -72,6 +72,10 b' Convert a foreign SCM repository to a Me'
     Mercurial Source
     -----------------
 
+    --config convert.hg.ignoreerrors=False (boolean) 
+        ignore integrity errors when reading. Use it to fix Mercurial
+        repositories with missing revlogs, by converting from and to
+        Mercurial.
     --config convert.hg.saverev=True          (boolean)
         allow target to preserve source revision ID
     --config convert.hg.startrev=0            (hg revision identifier)