Re: Jigdo feature request
On Mon, Dec 08, 2003 at 02:49:50PM +0000, Steve McIntyre wrote:
> At the moment, the jigdo images also pick up on various bits and pieces
> of READMEs that live in the root of the mirror, and various other docs.
> If these change between point releases (or if people are not keeping full
> mirrors), it's very common to have to download these files every time the
> user creates an image.
It would be nice if we could completely eliminate this problem, it causes a
lot of confusion for the users: "I just saw a 404 - surely the download is
There is a problem with jigdo-lite in that it downloads the complete
outdated files first, then realizes they have the wrong md5sums, and only
looks at the fallback mirror after trying to fetch lots of other files from
the first mirror. This is hard to change - have a look at the shell code in
jigdo-lite and you'll see what I mean...
The jigdo GUI will handle this more intelligently: It'll download the first
4k of the file and then match it against a checksum. If the checksum is
wrong, the next mirror is tried. Furthermore, this mirror is tried
immediately, rather than a long time after the first mirror; that way, the
probability that the user aborts the download in the meantime is low, and
at most very few files will be downloaded again when jigdo is restarted.
> It would be nice if jigdo could be told to always ignore (i.e. always
> place in the template file) files whose names matched a user-specified
> regexp. We could then specify --ignore=/README --ignore=/doc/ to kill
> these at build time.
> Or it might be a better move to remove these files from the list we
> generate in jigdofilelist. But I think the existing code in debian-cd
> looks like it tries to do this, and clearly it's not working.
IMHO then debian-cd needs further tweaking. I'm against adding "find"-like
features to jigdo-file, because it's always possible to run "find" and pipe
its output into jigdo-file.
|_) /| Richard Atterer | GnuPG key:
| \/¯| http://atterer.net | 0x888354F7
¯ '` ¯