[ 
https://issues.apache.org/jira/browse/MNG-7700?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17693385#comment-17693385
 ] 

Elliotte Rusty Harold commented on MNG-7700:
--------------------------------------------

The tricky cases all seem to involve leading zeroes. E.g. 0.7 resolves to 7, 
not to 0.7. This feels wrong to me, but it is the behavior of the code and I 
can see an argument for it in 
https://maven.apache.org/pom.html#version-order-specification.
I think we need to revise both code and spec so that 0.7 and 0-7 do not 
canonicalize to "7", perhaps by more carefully defining what we mean by a 
*trailing* null value.

> Improper canonicalization of versions
> -------------------------------------
>
>                 Key: MNG-7700
>                 URL: https://issues.apache.org/jira/browse/MNG-7700
>             Project: Maven
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 3.8.7, 3.9.0
>            Reporter: David M. Lloyd
>            Assignee: Elliotte Rusty Harold
>            Priority: Major
>             Fix For: 3.8.x-candidate, 3.9.1-candidate, waiting-for-feedback
>
>
> The canonicalization logic for versions is incorrect.
> Using the method {{ComparableVersion#getCanonical}} as in {{new 
> ComparableVersion(input).getCanonical()}}, the following results can be 
> observed:
> ||Input||3.8.6 Output||3.8.6 OK?||3.9.0 Output||3.9.0 OK?||
> |{{1}}|{{1}}|yes|{{1}}|yes|
> |{{0.1}}|{{0.1}}|yes|{{0.1}}|yes|
> |{{0-1}}|{{1}}|no|{{1}}|no|
> |{{1.x}}|{{1.x}}|yes|{{1-x}}|maybe*|
> |{{1-x}}|{{1-x}}|yes|{{1-x}}|yes|
> |{{0.x}}|{{0.x}}|yes|{{x}}|no|
> |{{0-x}}|{{x}}|no|{{x}}|no|
> |{{x}}|{{x}}|yes|{{x}}|yes|
> |{{0.rc}}|{{0.rc}}|yes|{{rc}}|no|
> The "OK?" columns indicate whether parsing the canonical version string will 
> yield a {{ComparableVersion}} instance that is {{equal}} to one constructed 
> from the original input, i.e. it's internally consistent.
> The "maybe*" item indicates that starting with 3.9.0, version `1.x` is now 
> considered to be equal to `1-x`. I'm not sure if this is a bug or not, or was 
> intentional or not, but it is definitely a change.
> These canonicalizations seem to have gotten less consistent in the move to 
> 3.9.0.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to