Comment 5 for bug 149113

This might have been broken by the path_content_summary changes. Rev 2879 changed commit to call this new routine. The code itself for path_content_summary was introduced in 2862. It needs more digging but the place that stands out as suspect is the implementation of path_content_summary in workingtree.py, particularly this bit:

            if not supports_executable():
                executable = None # caller can decide policy.
            else:
                mode = stat_result.st_mode
                executable = bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)

For win32 systems, this will always set executable to None. So either this bit of code is wrong, or the code calling it needs to do something intelligent when it gets back the combination of "kind is 'file' and executable is None".