Michael Haggerty <mhag...@alum.mit.edu> writes:

> Somebody who knows more about how diff operations are configured
> should please review this. I'm not certain that the change as
> implemented won't have other unwanted side-effects, though of course
> I checked that the test suite runs correctly.

Generally, I think this is a bad idea.  

We run "diff-tree" internally for multiple purposes:

 - When the same path we are drilling down appears, we use diff-tree
   to compare that path alone, to obtain textual diff, so that we
   can say "what did not appear in the textual diff output are
   attributable to the parent commit, we can exonerate this child".

   Even if the command line to "git blame" had "-Sfoo", you do not
   want to pass it down to this invocation.  If the child did not
   change the number of occurrences of string "foo" in the path
   being drilled down, it will pass down the blame for all lines to
   the parent.  And copying -R from the command line to this
   invocation does not make any sense.  Copying -C -C from the
   command line will defeat the whole purpose of having find_origin
   vs find_rename distinction, I would think, for this step.

 - When the path we have been drilling down for no longer appears in
   the parent, we use diff-tree with rename detection to find where
   the file _could_ have come from.

   It is a good idea to allow the similarity threshold to be tweaked
   from the command line.  Copying -R from the command line to this
   invocation may make sense, but I think you would break this step
   if you copied -C/-C -C

 - When -C/-C -C/-C -C -C is given from the command line, after
   running the "pass the blame to our primary suspect", we run
   tree-level diff-tree with find-copy option, only to enumerate
   paths that appear in the parent.  We pick pieces from these paths
   by doing blob-level diffs in diff_hunks() ourselves.  Copying -C
   from the command line would be useful if you are only doing a
   single '-C' (because it would allow you to tweak the similarity
   threshold), but otherwise it would probably break the command.

The design principle taken in "git blame" is to leave the decision
on what diff options do or do not make sense for these particular
invocations of the internal "diff-tree", and have "blame" give the
selected options to the internal "diff-tree" invocations.  "-w"
happens to use xdl_opts that will eventually be passed down to diff
machinery but you should consider it an accidental optimization
(i.e. "blame" designer considered that it makes sense to give
whitespace-ignoring comparison at the leaf-level "diff between
blobs", and found that the most efficient way to do so is to just
take "-w" from the command line and set the ignore-whitespace bit
used to drive the internal "diff-tree" invocation).  If we _were_
adding -R<num> to allow the users to affect similarity threshold,
we would parse it in "git blame" command line option processing,
and would give that _ONLY_ to the second invocation above, i.e. the
one done in find_rename() but not in others like in find_origin.





--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to