Hallo Joachim.
Qua, 2010-01-06 às 19:58 +0100, Joachim Breitner escreveu:
(...)
> Or we could cook up a patch to uscan so that
> http://hackage.haskell.org/packages/archive/darcs/([\d\.]*\d)/darcs-([\d\.]*\d).tar.gz
> actually works with --download-version.
This would be great, but isn't it going to break how other users expect
uscan to work? In the manpage it sais that in HTML only one page is
downloaded to look for the versions, and this is not what we want. I
took a look at uscan code, and it downloads the newest dir that matches
the pattern before the filename.
Devscripts developers:
I'll try to explain what we have in mind using darcs as an example. The
versions of all haskell libraries and applications that uses Hackage are
disposed in this way:
The newest version is:
http://hackage.haskell.org/packages/archive/darcs/2.3.1/darcs-2.3.1.tar.gz
The current one is:
http://hackage.haskell.org/packages/archive/darcs/2.3.0/darcs-2.3.0.tar.gz
When we use a watch file like the one Joachim suggested above, uscan
--report-status works, but not uscan --download-current-version don't.
So we currently are using some opts to make it work:
opts="downloadurlmangle=s|archive/([\w\d_-]+)/([\d\.]+)/|
archive/$1/$2/$1-$2.tar.gz|,filenamemangle=s|(.*)/$|darcs-$1.tar.gz|" \
http://hackage.haskell.org/packages/archive/darcs \
([\d\.]*\d)/
When I first read the documentation of uscan I thought the first attempt
should work, but after reading with more attention the documentation and
specially the code, I noticed that it only downloads one page, with the
newest dir that matches the regexp. Do you think it's a good idea to
make uscan downloads all the matching pages, and search for the current
version in all of them?
Is there a better workaround for this?
Auf Wiedersehen.
(...)
--
marcot
http://marcot.iaaeee.org/