After converting my repos to hashed inventory using darcs 2.0.0, I get
the following output from darcs check --verify-hashes:

darcs failed:  Hash failed hash: /home/mithrandi/code/Fusion/trunk/_darcs/inventories/0000000820-a7b262b6d7c6d5acc790c8ac075bc4dfef04051fcfb02c66c8c87b3a47fea734
-- 
mithrandi, i Ainil en-Balandor, a faer Ambar

mithrandi1: let's try to make a minimal test case out of this.  Can you identify
what patch is it (perhaps by examining the file named in the failure message
that you quote) and generate a repository containing that patch and little else?

* Zooko <bugs@darcs.net> [2008-04-22 18:35:03 -0000]:

> mithrandi1: let's try to make a minimal test case out of this.  Can you identify
> what patch is it (perhaps by examining the file named in the failure message
> that you quote) and generate a repository containing that patch and little else?

Maybe I'm not understanding the failure message correctly; I thought it
indicated that the inventory file itself had the incorrect hash?
-- 
mithrandi, i Ainil en-Balandor, a faer Ambar

mithrandi: Any chance you could attach the inventory file in question?

* Eric Kow <bugs@darcs.net> [2008-04-30 19:42:15 -0000]:

> mithrandi: Any chance you could attach the inventory file in question?

It seems I don't have the exact repository that I reported on before,
but an equivalent error happens in the version I do have now:

darcs failed:  Hash failed hash: /home/mithrandi/code/Fusion/trunk/_darcs/inventories/0000016680-0189c456ec9ada98141d31d7c7904c16fc67d192c1a6593012ad552c19c9595d

I'm attaching that inventory file.
-- 
mithrandi, i Ainil en-Balandor, a faer Ambar

Attachments

• 0000016680-0189c456ec9ada98141d31d7c7904c16fc67d192c1a6593012ad552c19c9595d

Any volunteers willing to look further into this?

Thanks and best regards
Thorkil

I piped Tristan's (mithrandi1) inventory file through gunzip | shasum -a 256 and
the hash matched.

So I don't have a good idea what we can do to get to the bottom of this.  In the
meantime, Tristan, do you still get a hash verification failure with the latest
darcs?  Maybe having the whole repository could give us some more clues?

In the meantime, I don't have a very good idea what bits of code actually
perform the hash check.  I wasn't able to find the string 'hash failed' in the
source code for 2.0.2.  I wonder how we're getting this message exactly. 
Perhaps somebody could look into this and clarify?

On Sat, Sep 5, 2009 at 1:00 AM, Eric Kow <bugs@darcs.net> wrote:
> So I don't have a good idea what we can do to get to the bottom of this.  In the
> meantime, Tristan, do you still get a hash verification failure with the latest
> darcs?  Maybe having the whole repository could give us some more clues?

The version I now have on my system is "2.3.0 (release)", which
doesn't appear to have "darcs check --verify-hashes" at all anymore.
Perhaps the code that implemented this was just buggy, but has since
been removed? I don't really understand what it was doing to begin
with, it just seemed like a useful thing to run as part of verifying
that everything went well with the conversion.

On Mon, Sep 07, 2009 at 11:02:30 +0200, Tristan Seligmann wrote:
> The version I now have on my system is "2.3.0 (release)", which
> doesn't appear to have "darcs check --verify-hashes" at all anymore.
> Perhaps the code that implemented this was just buggy, but has since
> been removed? I don't really understand what it was doing to begin
> with, it just seemed like a useful thing to run as part of verifying
> that everything went well with the conversion.

Thanks! It looks like the feature was indeed removed with this patch:

Tue Apr 29 18:23:03 CEST 2008  David Roundy <droundy@darcs.net>
  * remove unneeded --verify-hash flag (we always verify hashes).

Using darcs changes --tag . --from-p 'remove unneeded --verify'
I determined that this was likely removed in Darcs 2.0.2

Now the interesting thing is that we now claim that we're always
verifying these hashes.  I'm not clear on what always means concretely,
but the fact that you haven't gotten a complaint from darcs since makes
me think it's safe to mark this bug presumed-dead (I'd say resolved, but
we don't know what actually got fixed).