On Saturday 14 November 2009 20:55:16 Maxim Wexler wrote: > > redirect to a file, bash it into suitable shape with your Unix text tools > > of course, use said file as input to wget. > > > > > > -- > > alan dot mckinnon at gmail dot com > > Here > > http://www.gentoo-wiki.info/TIP_Gentoo_for_dialup_users > > I found this gem: > > emerge -fpu world | sort | uniq | sed '/\(^http\|^ftp\).*/!d;s/\ > .*$//g' > links.txt > > But something doesn't seem right. links.txt has 92 lines(I added the > ND switches) that all use only one URL, distfiles.gentoo.org, for each > package. It's 5.5k. But the raw command lists several URLs for each > package and it's gotta be ~200k. And if you read the article the wget > command is meant to skip the other URLs as soon as one instance of the > pkg has been downloaded: > > "With wget, just do: > > wget -i links.txt -nc > > Option -i tells wget to look inside links.txt for URLs of stuff to > download, option -nc tells it not to download it twice or thrice once > the file has been retrieved from a working URL." > > Am I missing something here?
The output of emerge -f lists ALL known mirrors and SRCs configured on the machine for each distfile. So you really only need to grab the first one listed, which is usually gentoo.org. If the file is not there, it is most unlikely to be anywhere else, with the exception of fetch-restricted packages. Those you would have to download manually anyway. When I was forced to use this same method, I just used cut on the output, using space as the delimiter. This seldom failed. I would not trust wget -nc. It has all kinds of implications including what to do with differing timestamps. wget -c is better. It completes partial downloads. -- alan dot mckinnon at gmail dot com
