Hi, I'm trying to unmask a package which has the following template: name-1.0_p20160101 It was possible to keyword such packages with the following line in the previous releases: =category/name-1.0_p2016* However, it stopped working with the latest stable portage. I am not able to find a relevant documentation. Could you help to clarify is that a bug? If not, how the asterisk sign can be used in this case? I really want to keywords such packages based on a year or month. Thank you.
(In reply to Anton Bolshakov from comment #0) > I am not able to find a relevant documentation. Could you help to clarify is > that a bug? It's a result of the "fix" for bug 560466, which cause the operator to only match on boundaries between version parts. > If not, how the asterisk sign can be used in this case? The closest thing would be >=category/name-1.0_p20160101 > I really want to keywords such packages based on a year or month. You could ask the ebuild maintainer to split the versions like _p2016_p0101 or something like that, so there's a boundary there.
interesting. First of all, the suggested _p2016_p0101 looks completely weird, so let's not go there. I think the main problem is that "*" operates with a string but it is getting applied against a number which is basically a hack. The bug #560466 changed that behaviour and it acts more like "~>" operator in ruby: Gentoo: (=2.3*) Ruby "~>2.3.0" "equal to 2.3.0 or greater than 2.3.0, but less than 2.4.0" However, it will not allow to specify ~>2.3.2 logic.
(In reply to Anton Bolshakov from comment #2) > I think the main problem is that "*" operates with a string "*" is only a mnemonic to shell globbing. It operates as the implementation defines it to operate. > but it is getting applied against a number which is basically a hack. Even worse, it is applied on a _version_ number which in general can be written as various different string representations. The previous implementation was a not officially documented huge list of hacks and exception and was not useful in many cases. I am very glad that this is now standardized officially (and in a manner appropriate for a version number instead of a string, even if it violates some expectation; but every agreement will violate some expectation). > Gentoo: (=2.3*) > Ruby "~>2.3.0" "equal to 2.3.0 or greater than 2.3.0, but less than 2.4.0" Not exactly: 2.3_alpha would also be matched, although it is lower than 2.3 (not to mention that 2.3.0 is _strictly_ greater than 2.3; even 2.3.0_alpha lies strictly in-between.)
