On Fri, Apr 9, 2010 at 11:12 AM, Christian Marillat <[email protected]> wrote: > James Vega <[email protected]> writes: > >> On Fri, Apr 9, 2010 at 10:33 AM, Christian Marillat <[email protected]> >> wrote: >>> James Vega <[email protected]> writes: >>>> On Fri, Apr 09, 2010 at 11:04:26AM +0200, Martin Zobel-Helas wrote: > > [...] > >>> Debian policy is for ... Debian We don't have control for upstream >>> filename. >>> >>> The error is after scanning the *upstream* filename on remote site. >> >> Which we sort *after* applying uversionmangle. Like I said, >> uversionmangle specifically exists to convert upstream's version into >> something that works for Debian. > > uscan was working before without aplying any uversionmangle rules...
Which is because “dpkg --compare-versions” doesn't perform any checks on whether the given version is a valid Debian version. I think that performing validity checking is appropriate because it helps ensure that upstream versions are mangled into something that sorts properly, thus making uscan work better. With the old behavior, 2.4 sorts lower than 2.4_previewX so uscan won't automatically download the proper release. I'll run a check on the lintian lab in the next couple days to see how many watch files are affected by the validity checking. If there are a significant number, I'll commit changes such that comparisons of upstream versions don't perform validity checking and comparisons of Debian versions do. Otherwise, I think the proper path forward is to use uversionmangle as this will cause sorting to be more reliable. Does this sound reasonable? -- James GPG Key: 1024D/61326D40 2003-09-02 James Vega <[email protected]> -- To unsubscribe, send mail to [email protected].
