Re: diff error? url@rev against local mods -- was: Re: diff_repos_wc question

Julian Foad wrote:
>>> Is this really an error or am I still missing something?
>
> It's really an error.
>
> Can you maybe add a test for it, and/or file an issue, and/or fix it, or
> just ignore this case, while still writing the directory compare thing?

Other than this error appearing in the tests, the directory compare thing is
already working. Your suggested way of implementing diff_summarize looks
very good in terms of avoiding to re-invent the wheel.

Now, if I/we could manage to fix the cases where diff is just broken, it
would be a nice side-effect of tree-conflicts. Is it a -- as stsp would say
-- huge can of worms, though? What other diff failures are known?

I'd still +1 on carrying on with the diff_summarize solution of dirs_same_p,
for now ignoring the breaking freak cases. It drastically raises the general
incentive to fix (and exposure of) broken diff cases.

Julian, your original patch suggestion contains a change to files_same_p(),
which I've dropped because I didn't understand it then. I'll still see
whether I'll re-include it:
* subversion/libsvn_client/merge.c
  (files_same_p): Change to an implementation more like we're going to use
    for directories - one that compares against the repository rather than
    a local temporary file - to help me get up to speed on how to do this.
  (merge_file_deleted): Use the new version of files_same_p().

A couple of tests still fail with my patch, I'll investigate.
merge_tests.py 3 88 89 114 115 119
diff_tests.py 45

~Neels

-- 
Neels Hofmeyr -- elego Software Solutions GmbH
Gustav-Meyer-Allee 25 / Gebäude 12, 13355 Berlin, Germany
phone: +49 30 23458696  mobile: +49 177 2345869  fax: +49 30 23458695
http://www.elegosoft.com | Geschäftsführer: Olaf Wagner | Sitz: Berlin
Handelsreg: Amtsgericht Charlottenburg HRB 77719 | USt-IdNr: DE163214194