Re: [gentoo-dev] Undefined cases in version policy

[Georgi Georgiev]

maillog: 28/02/2004-02:47:02(+0100): Marius Mauch types
> Hi there,
>
> as I was rewriting the code used for verifying and comparing versions I
> noticed a case that's not covered in the docs, let's use the example
> where foo has the following versions available:
> 	foo-1.0b
> 	foo-1.0.1
> Now the question is: Which version is considered newer?
> As I had no idea and though that we didn't have this case in the tree I
> started a small poll in #gentoo-dev and achieved a 4:0 vote for foo-1.0b
> being newer than foo-1.0.1, so I went on and implemented it this way.
> Now while testing I noticed that a) current code handles this the other
> direction and b) we actually have 2 packages that are affected by this:
> app-arch/arj and media-video/kavi2svcd.
>
> So I'm sending this mail to get some more opinions about this. If anyone
> is curious, the actual code and test scripts are on bug #37406.

If there are only two packages, why not implement it the way those packages
have implemented it? However, since this type of versioning makes no sense, I
would rather reversion what is available in portage. For example, if 1.0b is
supposed to be older than 1.0.1, I'd make 1.0b -> 1.0.0b

I'd personally vote for 1.0b < 1.0.1, because the letter (in my understanding)
is a suffix that goes after MAJOR, MINOR and PATCHLEVEL (and any other levels
that might be there).  Unless there is a chance of seeing foo-1.0b.1 I mean.

And the score is:

1.0b : 1.0.1
   4 : 1

-- 
(    Georgi Georgiev   (  You are as I am with You.                    (
 )    chutz@×××.net     )                                               )
(   [phone]   (                                               (