[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [taw1@nic.fi: Bug#71091: Aptitude: unable to install packages from local mirror]



On Fri, Sep 08, 2000 at 06:55:50PM -0600, Jason Gunthorpe <jgg@ualberta.ca> was heard to say:
> Now.. The acquire system is kind of stupid and that is where the problem
> must lay. The first thing that happens is that 'StoreFileName' (which is a
> reference to the FileNames array) is set to the string you see in the
> error message. Later, after download is complete that name is fixed up to
> be an absolute path. Since this is not happening the pkgAcqArchive::Done
> method must not be called, or must some how be confused (memory trashed?) 

  I hope I'm not stomping on memory..anyway, I'd expect to see random crashes
and other weirdness from that sort of bug (I don't)

> Off hand, the only normal way I could see this being broken is if your
> Pulse method returned false which would abort the downloading and leave
> things in the strange state you see. If you trap the Cancelled result from
> Run you should be able to eliminate that possibility.

  I test the result of Run to be sure that it's "Continue".  (although I've
also been seeing annoying error messages, "Tried to dequeue a fetching item",
when I cancel a download; perhaps that's related?  I haven't done anything
about them yet because they appear to be harmless..)

> Try testing your code on auric with its local archive?

  I can't, but I did make a (minimal) local archive on my system and the bug
popped up immediately.

  Daniel

-- 
/----------------- Daniel Burrows <Daniel_Burrows@brown.edu> -----------------\
|      The sigfile hits!     |      Apostrophes are not a warning that a      |
|     You feel confused.     |      word is about to end in an "s".           |
\-------- Real Programmers don't have braces. -- http://www.python.org -------/



Reply to: