nirenjan / vmlib Goto Github PK

According to SemVer 2.0.0, leading zeroes are acceptable in build identifiers. However, current version of vmlib strips any leading zeroes and stores the identifiers as numbers (if the identifier has only digits).

There is nothing in format to prevent the user from passing in a string such as %3M.%2m.%p%M

In this case, the format code assumes that there is only 1 %(\d*)M element, and uses that to determine the format for the entire string. So while the user may be expecting 001.02.3.1, he would actually get 001.02.3.001 (note the final 00 padding).

The vmlib wrapper script serves to automate some of the common tasks implemented by the VMLib module. The issue is that if it prints out any help, the entire path is printed out, such as the line below

Usage: /Users/nirenjan/.rvm/gems/ruby-1.9.3-p429/bin/vmlib <command> [arguments]

What we would expect is to strip out the path and simply print

Usage: vmlib <command> [arguments]

As a result of this bug, you can see the full path being printed everywhere, and it throws off the formatting.

Based on SemVer v2.0.0, prerelease identifiers that are numeric MUST NOT contain leading zeroes. Currently, vmlib accepts a prerelease identifier of the form 01.002.0003, but stores it as 1.2.3 - which is compliant with the specification.

However, there is a corner case if the user expects to retrieve the number exactly as stored. There are three possible ways to address this.

1. Ignore leading zeroes in numeric identifiers. They will be stored as is, if and only if, there are any alphabetic characters or hyphens in the identifier. This corresponds to current behaviour.
2. If an identifier has leading zeroes, then treat it as a string.
3. Throw an error if the user attempts to enter a numeric identifier with leading zeroes.