[Cado-nfs-discuss] An old friend.

Paul Leyland paul.leyland at gmail.com
Tue Aug 19 10:09:41 CEST 2014


I rebuilt the latest repository version of cado-nfs on my Gentoo boxes.
After a week running on 14 cores a merging phase started and abruptly
finished with complaints about corrupted .gz files.  I tried to recover
but gave up.

It looks like you need to apply this patch to

8<----------------------------------------------------------------->8
*** wuclient2.py~	2014-08-10 14:41:40.626170875 +0100
--- wuclient2.py	2014-08-17 16:35:51.771305147 +0100
***************
*** 136,142 ****
  # These Python version have bug type #2
  BUGGY_MIMEENCODER2 = (
      (3,2,4), (3,2,5),
!     (3,3,2), (3,3,3), (3,3,4)
  )
  
  HAVE_WGET = False
--- 136,142 ----
  # These Python version have bug type #2
  BUGGY_MIMEENCODER2 = (
      (3,2,4), (3,2,5),
!     (3,3,2), (3,3,3), (3,3,4), (3,3,5)
  )
  
  HAVE_WGET = False
8<----------------------------------------------------------------->8

Ho hum.  Can this be fixed permanently, please, rather than have to
remember to re-patch every time python3 is upgraded?

A little script which automates recovery would be nice if the permanent
fix isn't feasible.  The script would use the database to identify
relations files (as it now does), check their integrity.  Those found to
be corrupt would be removed from the DB and the filesystem (including
any stderr logs) and the now missing tasks made available for
allocation.

With the patch installed a follow-up and easier factorization is now
merging.  Lesson learned the hard way so only three 14-core days at
risk.


Paul




More information about the Cado-nfs-discuss mailing list