I'm seeing bad output from annotate, with unidiff markers included,
sometimes with multiple lines printed for each entry.  In the example
below the file does not have any leading `-'s, c.f. the second entry.

I see this both with Ubuntu's 2.3.0 and the current source (`2.4 (+ 171
patches)'), and with or without my normal --diff-opts -u default.  I
could supply a (220K) tarball of the repo.

Attachments

• unnamed

Hi Dave,

Is it the case that this is incorrect output in your opinion or just not
helpful output?

I'm asking this because (I confess) I don't really use darcs annotate
very much so I don't have a clear idea what to expect.  It seems
plausible to me that darcs annotate could deliberately coalesce lines
that are identified with the same patch.

If I understand correctly, this is the darcssum repo?  Maybe a link and
a darcs annotate command line could help.  I couldn't find the link
after a couple minutes of Googling (I also recall Simon Michael has
taken it over).

Any chance you'd be able to see if you get the same sort of results with
darcs 1.0.9?  [that's a somewhat unreasonable request on my part, feel
free to disregard]

Eric Kow <bugs@darcs.net> writes:

> Eric Kow <kowey@darcs.net> added the comment:
>
> Hi Dave,
>
> Is it the case that this is incorrect output in your opinion or just not
> helpful output?

I think it's incorrect, but I don't have a specification of the output.
It's surely incorrect to say `line added by' and then splurge several
lines which aren't what was added (even subtractions rather than
additions).

> If I understand correctly, this is the darcssum repo?  Maybe a link and
> a darcs annotate command line could help.  I couldn't find the link
> after a couple minutes of Googling (I also recall Simon Michael has
> taken it over).

Yes.  I'll see if I can provide a recipe on the basis of the repo I have
at home, but I think the output was from an un-sent patch.

> Any chance you'd be able to see if you get the same sort of results with
> darcs 1.0.9?  [that's a somewhat unreasonable request on my part, feel
> free to disregard]

I guess so, if it will build with a current ghc installation, but the
relevant repo is hashed, so I suppose that's out.  I probably did the
vc-darcs.el annotate support with darcs 1, but I don't remember if I
looked at the code for the format or just experimented with the output
in the relevant cases.  (Obviously I assumed initially that I'd cocked
up vc-darcs.)  I haven't looked at the current Darcs code, but I don't
know my way around it, so it would be relatively hard work.

This obviously isn't a very important issue.

Eric Kow <bugs@darcs.net> writes:

> Is it the case that this is incorrect output in your opinion or just not
> helpful output?

Maybe I should have said that it doesn't seem an ideal format for the
annotate output as it's horribly verbose, and it would be better at
least to amalgamate runs of lines from the same patch, but I guess that
would be a separate wishlist issue.

Benedikt pointed out that the initial (+/-) that Dave noticed is
probably NOT normal (which seems to match my hazy memories).

Also, the link in question is http://joyful.com/repos/darcsum

He says he'll try to reproduce it.  Thanks, all!

Eric Kow <bugs@darcs.net> writes:

> Eric Kow <kowey@darcs.net> added the comment:
>
> Benedikt pointed out that the initial (+/-) that Dave noticed is
> probably NOT normal (which seems to match my hazy memories).
>
> Also, the link in question is http://joyful.com/repos/darcsum
>
> He says he'll try to reproduce it.  Thanks, all!

Just pulling that repo doesn't demonstrate it, and I've lost the context
that yielded the original samples, but my current repo does demonstrate
an instance.  However, it does seem as if the format is intentional on
the basis of issue 25.  In that case it could surely do with some doc,
if only to point out the problem with the format.  I'm not sure I
understand it to contribute that.

The --xml output doesn't have this problem, but I baulk at dealing with
it, especially as it doesn't seem to be defined, or even necessarily
valid.

I think we can close this as darcs has now a completely different (and
extremely more useful) annotate format, which most probably means that
this error can no longer happen.