Hi, Quoting Michael Stapelberg (2015-10-18 23:10:06) > I can’t figure out how to specify multiple binary packages when calling > dose-ceve. The manpage for -r says: > > -r pkgspec > Using the same syntax as in -c, this option use the reverse > dependency relation to make the transitive closure. > This option can also be specified as --rcone=pkgspec. > > So the manpage for -c says: > > -c pkgspec > The match of an atomic dependency (a package name p possibly > together with a version constraint c) is the set of all packages in the > repository with name p, and a version that satisfies the constraint c. The > dependency cone of a package p is the set of all matches of all atomic > dependencies of p, together with their respective dependency cones. The > package specification pkgspec is a list of packages (separated by a > semicolon), where each package is specified as follows: (name,version).
Right, unfortunately it turns out that when we recently implemented the feature that the -c and -r option would take a string in the Debian dependency format when handling Debian packages, we forgot to document it :( > Now I tried specifying multiple binary packages, but couldn’t get any > combination to work: > > https://paste.debian.net/316796/ (so as to not make this email too long) > > So, I’m at a loss. What am I misunderstanding here? Can you please provide > an example invocation of how you think ratt should call dose-ceve in this > specific case? The following will work: dose-ceve --deb-native-arch=amd64 -T debsrc -r 'golang-golang-x-tools (= 1:0.0~git20150716.0.87156cb+dfsg1-4),golang-golang-x-tools-dev' -G pkg deb://Packages debsrc://Sources You can see that the format of the -r option is the same as in a Depends field. > Also, may I suggest the following improvements to dose-ceve: > > 1. When -r is specified multiple times, it should not overwrite the package > spec, but amend it. If you think -r should only be specified exactly once, > I suggest dose-ceve should error out when users specify multiple -r values. > > 2. The manpage ceve(1) should come with an example for pkgspec. > > 3. Instead of merely stating that the provided pkgspec is invalid, > dose-ceve should tell the user why the pkgspec is invalid, and ideally > include a valid example. > > (4. Possibly, the manpage ceve(1) should be worded a bit more clearly with > regards to pkgspec, but perhaps it’s just me…) No it's not just you and I agree with all your suggestions. To not forget about your valuable input, I submitted a bug to the upstream bugtracker: https://gforge.inria.fr/tracker/index.php?func=detail&aid=19616&group_id=4395&atid=13808 Thanks! cheers, josch
signature.asc
Description: signature
