[MERGE] Fix for 306394: warning, not error, on non-existent file.

[Karl Fogel]

Aaron Bentley <aaron at aaronbentley.com> writes:
> The two parts of my critique are:
> 1. If non-existence is a status, it should go to stdout, instead of
> being a warning and going to stderr.
> 2. If non-existence goes to stdout, it should match our existing output
> styles.
>
> However, in the supplied patch, Karl invokes trace.warning, so it fails
> 1.  He stringifies a PathsDoNotExist, but this does not match either our
> default status output or our short status output.  So it fails 2.
>
> So as you can see, his proposed patch does not follow my critique.  You
> seem to agree that my critique is consistent with his argument,
> therefore his patch is inconsistent with his argument.  That is the
> point I was trying to make.

Makes sense to me.  I can do two things now:

   1) Redo the patch so the message goes to stdout and matches existing
      output styles,  -OR-

   2) Redo the patch so that non-existent files are silently skipped.

We haven't talked much about option (2).  I raise it now because it's
what Subversion does, and it seems to work out okay.  Although I think
it's a reasonable way to behave, it's unrelated to the logic behind (1).

My main concern is that non-existence of one thing not interfere with
status being reported for other, existent things.  Both (1) and (2) seem
fine to me; I'm not sure there's One Right Answer here.

Do people have strong preferences for (1) or (2)?  I'll wait for
feedback before writing a new patch.

-Karl