Re: ftpsync test please
Carlos Carvalho wrote:
> Raphael Geissert (firstname.lastname@example.org) wrote on 8 May 2012 01:48:
> So you have to clearly specify the new format. For the upstream, we
> can just write a line with the configured value, right? Note that it
> may be the IP or a different host name than what's published. It may
> even be a non-ascii name, encoded or not... But I'm fine with this if
> it's useful to you.
> About the architecture list,
> >VALUE := "FULL" | TOKEN [" " ARCH_LIST] | ARCH_LIST
> FULL is easy, no problem. But if the mirror wants to have
> ARCH_EXCLUDED, to construct the local archs we need the list of all
> architectures in the repository. Also, if the admin uses
> ARCH_LIST=FULL to not have to list every one, the script must know it.
Like Joerg said in another message: for ftpsync, we plan to switch from
"include all but some" (i.e. exclusions-based) to "exclude all but some"
(i.e. inclusions-based). As soon as we move to that, there's no need to know
the list of architectures in the archive.
For FULL, you simply include everything. Consider the following call to
rsync $arch_opts --exclude '*'
Then, for FULL you just have to set arch_opts to "--include '*'"
> >ARCH := "amd64" | "i386" | ...
> So you have to spell the full list, without ellipsis. Even if we know
> them, we need to know how you recognize them, if case matters, etc.
If you adhere to the definiton I gave you, you have the answer already.
To spell it out: they should be in lower case.
> However, we cannot hard code the full list in the script because it
> changes from time to time. Thus it's necessary that the admin puts it
> in the config, which is quite inconvenient. There must be a way for
> the script to discover it. What if a scripts takes the arch values
> from indices/files/arch-*.files of the master (not the local ones)?
There's no need for that, see above. You shouldn't trust your upstream's
> Also, what's the precise line format for each entry? And which order
> in the file?
1. $(LC_ALL=POSIX date -u)
2. Your program's signature
3. "Running on host: $(hostname -f)"
4. "Architectures: "...
5. "Upstream-mirror: $upstream_host"
n.b. IIRC the upstream mirror field was added to be able to generate the
mirrors graph without recurring to scanning/guessing. It is not used by
http.d.n, at least for now.
Raphael Geissert - Debian Developer
www.debian.org - get.debian.net