[darcs-users] unapplicable patch failure

Will will at glozer.net
Wed Jul 28 05:21:38 UTC 2004

David Roundy <droundy at abridgegame.org> writes:

> On Mon, Jul 26, 2004 at 09:30:57PM -0700, Will wrote:
>> Hello,
>> The following repository gives a "darcs failed: Unapplicable patch!"
>> error when getting, checking, or repairing it.
>>   http://www.newartisans.com/johnw/darcs/ledger2/_darcs.tar.bz2
>> The failed patch just contains some code hunks, but the changed file
>> is removed by a later patch.
> I see the problem, but it's hard to see how it happened.  Unfortunately
> it's four days older than the most recent change in the repository, so the
> odds of you remembering how you triggered it (assuming it's a darcs bug)
> are probably pretty slim.  :(

Thanks for checking David!  Unfortunately this isn't my repo but a
friend's and he doesn't remember the details of what happened.

> The problem is the
> Sun Jul 18 02:44:59 EDT 2004  johnw at newartisans.com
>   * add some new code, but commented it ,out
> (as I'm sure you noticed).
> This patch contains the following change:
> hunk ./register.cc 68
> -  bool          revalues = price_revalues(format.total_style);
> -  unsigned int  index    = 1;
> +  unsigned int  index    = 0;
> which is problematic.  There are probably other problem changes in there,
> but this is the one that I looked at.
> A grep of all your changes for price_revalues shows that this line with
> bool price_revalues only shows up in two patches, this one and a later
> one.  So obviously there is a "missing" change which needs to come before
> this one, in which the
>   bool          revalues = price_revalues(format.total_style);
> line was introduced to the file.  Somehow it seems that _darcs/current got
> messed up...
> Could you perhaps have hit ctrl-C or run kill during a record?  ctrl-C
> *shouldn't* cause a problem, but if you killed darcs during a record (or
> unrecord, etc) (or it got killed by the OOM killer or something), and it
> left a lock file that you then deleted, that could explain things.
> Probably any time a lock file is left over you should run a darcs check
> (followed by a repair if necesary).  I'm very interested in figuring out
> how this happened.

This could have happened, wish I could give you some more information
about the repo.  I'll relay the recommendation to always run a check.

> As far as what to do now to fix things, if you don't mind losing a few days
> of history, the easy fix is to unrecord all the changes back to before this
> patch, then run darcs repair (and darcs check to make sure the repair
> worked), and then record again the changes that you unrecorded (perhaps in
> one big change).

That sounds like a good plan, thanks!


More information about the darcs-users mailing list