[darcs-users] darcs patch: added regression test for issue1632

[Benjamin Franksen]

I'll amend and re-send (that's why I asked for review).

Trent W. Buck wrote:
> benjamin.franksen at bessy.de writes:
>> +## Test for issue1632 - <SYNOPSIS: 'darcs changes
>> dir/file-that-never-existed' +## should not list any patches.>
> 
> The surrounding <SYNOPSIS: > part isn't needed.  

Ok (this wasn't clear from the example).

> I would probably also
> clarify that dir *does* exist (as far as pristine is concerned), e.g.
> 
>     darcs changes D/f should not list any changes, where D is part of
>     the repo and f is a non-existent file.
> 
> ...since I'm assuming that "adddir dir/" must be recorded in order to
> trigger this bug.

Yes and yes.

>> [...]
>> +# This one lists no patches:
>> +darcs changes non-existent-file | not grep 'added dir'
>> +# But this one lists the dir creation patch:
>> +darcs changes dir/non-existent-file | not grep 'added dir'
> 
> I would prefer these comments to describe what *should* happen, rather
> than the current behaviour, i.e.
> 
>     # Darcs should not list any changes here.
>     # Darcs should list the dir creation patch here.

These would not be correct but I get your meaning. You want to just rename 
the script after the fix and not have to meddle with the comments, right?

> I would also redirect output (>log) so as to test Darcs' exit status,
> e.g.
> 
>     not darcs changes dir/non-existent-file >log
>     not fgrep 'added dir' log

What is the advantage of this?

BTW, is there any documentation about darcs return codes? (Asking because 
you wrote 'not darcs changes ...'.)

Cheers
Ben