darcs

Issue 1838 darcs failed: Couldn't fetch {HASH} in subdir pristine.hashed (noslurps)

Title darcs failed: Couldn't fetch {HASH} in subdir pristine.hashed (noslurps)
Priority bug Status resolved
Milestone 2.5.0 Resolved in 2.5.0
Superseder Nosy List dmitry.kurochkin, gh, kowey, mornfall
Assigned To mornfall
Topics Hashed, Regression

Created on 2010-05-06.13:21:09 by kowey, last changed 2010-06-28.16:57:40 by mornfall.

Messages
msg10978 (view) Author: kowey Date: 2010-05-06.13:21:08
darcs get http://code.haskell.org/wxhaskell succeeds with darcs 2.4.1
but with current HEAD (just after noslurps), I get:

darcs failed:  Couldn't fetch
`696806fc8aa4d5e5371a1ba8d82e7611a306d97a24105ccaa2712712e41d7d0b'
in subdir pristine.hashed from sources:

This is converting from an old-fashioned repository

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
msg10979 (view) Author: kowey Date: 2010-05-06.13:22:18
Needs investigation.  I'm sure Petr will work out what's wrong :-)
Hopefully we can boil this down into a minimal test case.
msg10982 (view) Author: gh Date: 2010-05-06.14:21:14
Ono my computer, with current HEAD just after noslurps also, I don't
have this error but darcs simply freezes. straces tells me only this:

$ strace -p `pidof darcs`
Process 23383 attached - interrupt to quit
futex(0xa34ab70, FUTEX_WAIT_PRIVATE, 9, NULL

This was with a cold cache.
msg10994 (view) Author: gh Date: 2010-05-07.11:04:09
By "freezing" I mean "hanging", ie, that darcs was displaying nothing
and seemed to be doing notinhg at all (as reflected by strace).

This behaviour was obtained by doing a "darcs get
http://code.haskell.org/wxhaskell "
msg11001 (view) Author: mornfall Date: 2010-05-07.15:39:50
Hi,

gh <bugs@darcs.net> writes:
> By "freezing" I mean "hanging", ie, that darcs was displaying nothing
> and seemed to be doing notinhg at all (as reflected by strace).
>
> This behaviour was obtained by doing a "darcs get
> http://code.haskell.org/wxhaskell "
I had the same thing yesterday on code.haskell.org and turned out wget
hangs just as well. So I would disregard this as just deaf HTTP on
code.haskell.org (arguably, we should print *something*...).

Yours,
   Petr.
msg11024 (view) Author: mornfall Date: 2010-05-08.09:19:12
Eric, could you try reproducing with just local repositories if you have 
a moment? Thanks... (Of course, as always, a test would be best.)
msg11030 (view) Author: gh Date: 2010-05-09.10:22:30
mornfall> You were right. I've just tried again with HEAD compiled with
cabal install -fstatic -f-curl , I now have the same problem than Eric.
msg11043 (view) Author: kowey Date: 2010-05-11.16:06:56
On Sat, May 08, 2010 at 09:19:13 +0000, Petr Ročkai wrote:
> Eric, could you try reproducing with just local repositories if you have 
> a moment? Thanks... (Of course, as always, a test would be best.)

Still fails! (which seems like a good thing)

darcs get --old-fashioned http://code.haskell.org/wxhaskell
darcs get wxhaskell wxhaskell2

darcs failed:  Couldn't fetch
`696806fc8aa4d5e5371a1ba8d82e7611a306d97a24105ccaa2712712e41d7d0b'
in subdir pristine.hashed from sources:

thisrepo:/tmp/wxhaskell-2
cache:/home/kowey/.darcs/cache

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
msg11617 (view) Author: mornfall Date: 2010-06-28.14:41:53
With "just local" I meant ruling out http out of the equations. Ta.
msg11618 (view) Author: kowey Date: 2010-06-28.14:45:17
On Mon, Jun 28, 2010 at 14:41:54 +0000, Petr Ročkai wrote:
> With "just local" I meant ruling out http out of the equations. Ta.

Yep! This was buried under the first line of my test :-)

darcs get wxhaskell wxhaskell2

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
msg11623 (view) Author: mornfall Date: 2010-06-28.15:14:22
Hmm!

$ darcs get --old-fashioned http://code.haskell.org/wxhaskell
Official wxHaskell darcs repository
**********************
Finished getting.                    

$ darcs get wxhaskell wxhaskell-hashed
Converting old-fashioned repository to hashed format...
*******************************************************************************
Fetching a hashed repository would be faster.  Perhaps you could persuade
the maintainer to run darcs optimize --upgrade with darcs 2.4.0 or higher?
*******************************************************************************
Finished getting.

... am I missing something? (this is reasonably recent HEAD)
msg11625 (view) Author: kowey Date: 2010-06-28.16:32:47
As requested:

with HEAD, get completes just fine

with current HEAD minus the hashed-storage 0.5.2 bump:
darcs: Codec.Compression.Zlib: premature end of compressed stream

we get the same result with an attempt to an attempt to build a
Darcs resembling head at the time I reported the patch (which seems
tricky without a context file, sigh)

Petr: shall we mark this resolved then?

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
msg11626 (view) Author: mornfall Date: 2010-06-28.16:57:39
I agree, this is very likely to be the rename bug that's been fixed in 
hashed-storage 0.5.2.
History
Date User Action Args
2010-05-06 13:21:09koweycreate
2010-05-06 13:22:20koweysetstatus: unknown -> needs-reproduction
topic: + Regression, Target-2.5, Hashed
nosy: + mornfall
messages: + msg10979
priority: bug
assignedto: mornfall
2010-05-06 14:20:18ghsetnosy: + gh
messages: + msg10981
2010-05-06 14:21:00ghsetmessages: - msg10981
2010-05-06 14:21:15ghsetmessages: + msg10982
2010-05-07 11:04:09ghsetmessages: + msg10994
2010-05-07 11:04:12ghsetmessages: + msg10995
2010-05-07 11:04:50ghsetmessages: - msg10995
2010-05-07 15:39:51mornfallsetmessages: + msg11001
2010-05-08 09:19:13mornfallsetmessages: + msg11024
2010-05-09 10:22:31ghsetmessages: + msg11030
2010-05-11 16:06:57koweysetmessages: + msg11043
2010-06-15 20:52:14adminsetmilestone: 2.5.0
2010-06-15 21:00:01adminsettopic: - Target-2.5
2010-06-28 14:41:54mornfallsetmessages: + msg11617
2010-06-28 14:45:18koweysetmessages: + msg11618
2010-06-28 15:14:22mornfallsetmessages: + msg11623
2010-06-28 16:32:48koweysetmessages: + msg11625
2010-06-28 16:57:40mornfallsetstatus: needs-reproduction -> resolved
messages: + msg11626
resolvedin: 2.5.0