[darcs-users] Data loss bug when renaming and replacing a non standard token?

Eric Kow kowey at darcs.net
Sun Mar 22 21:52:01 UTC 2009


On Sun, Mar 22, 2009 at 19:11:55 +0000, Ashley Moran wrote:
>> When you say data loss, are you referring to stuff in his/her working
>> directory that was lost?
>
> Yes, he lost the text he added at the end of the file.  From my example, 
> it's equivalent to:

Argh!  Sorry about that.  If it's any consolation, I'm relieved it's
'just' working directory lost, although I understand that this can be a
pretty big deal :-(

> Think I understand you - the data had not been recorded, so it was truly 
> lost.  I assume an un-pull could not recover data in this case?

That's right.  I was hoping it was just an unexpected result (sorry,
I haven't taken the time yet to deeply understand the run-through)

>> So to make sure I'm on the same page: are you referring specifically
>> to the failure to back up a file in the presence of a conflict?
>
> Yes.  I would possibly have to double check I've remembered what I did  
> (I've closed the terminal window), but I've simulated it without a  
> replace operation, and with a replace operation that uses standard  
> tokens, and with a replace operation that doesn't need forcing.

Thanks!

>> Please submit a report to bugs at darcs.net, and also create a test case
>> for our bugs/ directory and darcs send it (since you *almost* have  
>> one!
>> <snip>

> Hmm, can I forward the original email to that address?

I expect that to work.

> I'd happy to create a test case and examine all the possible  
> permutations, but right now I have barely a minute spare, and it could  
> be a couple of weeks before I get round to it.  At a risk of sounding  
> lazy, can I leave it with the darcs team for now, and if nobody gets  
> chance to look at it, pick it up when I'm less busy?

Ooh.  Well I know that you have a system to keep you on top of things,
so please do take your time.

I would advocate just starting by submitting what you have as a test
case (it's pretty simple; the key to the test cases is [a] they are
shell scripts and [b] any non-zero exit code is treated as a failure
unless explicitly accounted for).

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: <http://lists.osuosl.org/pipermail/darcs-users/attachments/20090322/a2fb7cfd/attachment.pgp>


More information about the darcs-users mailing list