3

When I push a patch to a remote repository via ssh, the first try fails with 

me@bowler$ darcs push
~/Dokumente/Aufsaetze/ar_report
Pushing to "me@somewhere:/home/me/darcs_repos/ar_report"...
Mon Nov 17 10:13:38 CET 2014 me@somewhere.com
  * automatisierung
Shall I push this patch? (1/1)  [ynW...], or ? for more options: a

darcs failed:  Couldn't fetch 0000000369b60776b9ae996a253f8a5a296927293faab87036846b34b0e710c9fc0f58429b'
in subdir inventories from sources:

thisrepo:/usr/home/me/darcs_repos/ar_report
cache:/home/me/.darcs/cache


HINT: I could not reach the following repositories:
    /home/me/.darcs/cache
    /usr/home/me/darcs_repos/ar_report
  If you're not using them, you should probably delete
  the corresponding entries from _darcs/prefs/sources.
Apply failed!

The second attempt works without problems. Does someone has a clue what's going on here?

Steffen
  • 733
  • 2
  • 10
  • 24
  • What's in `_darcs/prefs/sources` and how did it end up that way? – tripleee Nov 17 '14 at 12:38
  • repo:me@somewhere:/home/me/darcs_repos/ar_report. Came with the get. – Steffen Nov 17 '14 at 12:44
  • 1
    The reference to `/usr/home` looks fishy. Do you have, or did you use to have, a symlink there? – tripleee Nov 17 '14 at 12:48
  • 1
    This could be the point. It is a FreeBSD box, where the old directory style was /usr/home/USER. I don't have this on my box, but there might be some legacy code. I'll go after that. – Steffen Nov 17 '14 at 12:53
  • correction: /home is a symlink to /usr/home. I changed the sources file to /usr/home/me/darcs_repos/ar_report. Same problem. Maybe a cache file somewhere? – Steffen Nov 17 '14 at 13:08
  • 1
    Does this happen for each new patch you try to push? Was either the local or the remote repository made with --lazy (or ^Cing a get)? – Ganesh Sittampalam Nov 17 '14 at 18:54
  • @ganesh-sittampalam It happens for each new patch. No lazy-get and no interrupted get. I just tested it. I am not sure, but it could be a problem with the symlink. The thing is: Why does the second attempt succeed? – Steffen Nov 19 '14 at 06:39
  • The second attempt succeeding is definitely the strangest aspect. Can you reproduce it if you use `darcs send -O` to write out a `.dpatch` file, copy it to the remote side manually and use `darcs apply` on the file? As the failure is at the remote end, if that does reproduce it, it'll make it simpler to debug. – Ganesh Sittampalam Nov 19 '14 at 07:47
  • This reproduces the failure. Second attempt succeeds. Should I file a bug report? – Steffen Nov 19 '14 at 09:28
  • Yes, go ahead. Can you include the output of both `darcs apply` commands with `--debug`? – Ganesh Sittampalam Nov 19 '14 at 16:58
  • Done. Bug report filed. For reference: http://bugs.darcs.net/ – Steffen Nov 20 '14 at 11:19
  • I can't actually see this bug there, and your link points to the front page. Did it get submitted properly? – Ganesh Sittampalam Nov 25 '14 at 16:04
  • I submitted it via email on 11/19 – Steffen Nov 26 '14 at 16:32
  • Unfortunately it seems to have got lost - we've checked the moderation system but can't find it :-( Would you mind resending, and copy ganesh@earth.li in case it goes wrong again? – Ganesh Sittampalam Nov 28 '14 at 22:29
  • Yes, sure! and... done. – Steffen Dec 01 '14 at 06:28
  • 2
    For reference, link to the issue: http://bugs.darcs.net/issue2424 – Steffen Dec 01 '14 at 06:41

1 Answers1

3

After some investigation, this turns out to have gone away in darcs 2.8.5 whereas it was happening in darcs 2.8.4. My best guess is that it was an instance of this bug, but the connection wasn't obvious.

Ganesh Sittampalam
  • 28,821
  • 4
  • 79
  • 98