submodule handling

[anonymous]

the code breaks at some places in git_fs, on commits having submodule changes. the first one i found is get_content_length returning None, and then trac trying to sum Nones. a quick workaround is changing

if not self.isfile:
  return None

to

if not self.isfile:
  return 0

the second is in get_changes. submodule changes have a different mode than directory changes, so they end up with kind = file, and trac tries to read the content. my workaround is changing the condition

if mode2.startswith('04') or mode1.startswith('04'):
  kind = Node.DIRECTORY

to

if mode2.startswith('04') or mode1.startswith('04') or mode1.startswith('16'):
  kind = Node.DIRECTORY

so these changes are treated like directory changes.

these are just some quick hacks, but they make the plugin usable on repos with submodules, until there's a better solution.

[anonymous]

Replying to anonymous:

the code breaks at some places in git_fs, on commits having submodule changes. the first one i found is get_content_length returning None, and then trac trying to sum Nones. a quick workaround is changing

if not self.isfile:
  return None

to

if not self.isfile:
  return 0

the second is in get_changes. submodule changes have a different mode than directory changes, so they end up with kind = file, and trac tries to read the content. my workaround is changing the condition

if mode2.startswith('04') or mode1.startswith('04'):
  kind = Node.DIRECTORY

to

if mode2.startswith('04') or mode1.startswith('04') or mode1.startswith('16'):
  kind = Node.DIRECTORY

so these changes are treated like directory changes.

these are just some quick hacks, but they make the plugin usable on repos with submodules, until there's a better solution.

[anonymous]

see also #3104

[anonymous]

Does anyone know why this was closed as duplicate? Applying the diff from 3104 doesn't fix the issue above, though it does address a different issue with submodules.