[MERGE][#175520] use historical context for file logging

[Ian Clatworthy]

Aaron Bentley wrote:
> Ian Clatworthy wrote:
>> Ian Clatworthy wrote:
>>> This patch fixes log so that the end revision is used
>>> for mapping a FILE to a file-id. This means log can
>>> now find deleted files. It also means renamed files
>>> are matched against their name at the time.
>> This updated version searches the start revision if the
>> file doesn't exist in the end revision.
> 
> This looks like a partial reimplementation of Tree.paths2ids.

Thanks for the pointer -I had forgotten about that API.

I think I want something more like _find_ids_across_trees()
though as I don't want children of directories. (Semantically
logging a directory would be the same as logging each file.)

I'm also curious about the overhead of creating the extra
trees, e.g. b.repository.RevisionTree(branch.get_rev_id(1)).
Currently, I only do that if the fileid wasn't found in
the end tree (typically the working or basis tree).

What would you recommend? Keep the current patch or
build all potential trees then call _find_id_across_trees()?

Ian C.