Find minimum required module version

lima1 has asked for the wisdom of the Perl Monks concerning the following question:

Hi,

did anyone wrote a test-script that tries to find the minimal version numbers for the modules listed in Build.PL or Makefile.PL? Probably something that fetches old versions from http://backpan.cpan.org and runs the test suite with each of them?

Comment on Find minimum required module version

Replies are listed 'Best First'.
Re: Find minimum required module version by davidrw (Prior) on Dec 08, 2008 at 17:43 UTC
This might be a good question to cross-post to the cpan-testers mail list, too. Some initial thoughts on it: What do you test again? So your module Z requires A, B, C. Do you prep w/the newest of B & C, and then try all versions of A (newest->oldest), running the test suite Z until it breaks? This would definitely require pretty decent coverage by Z's test suite. Might there be recursive problems with this? Maybe Z works fine with any B ... but A requires (for functionality that Z doesn't use; but still a requirement for install) a certain version of B. How does that get handled? Maybe such a script, instead of finding the min versions for each prereq independently, finds a point in time that they work? Construct a time line of release dates like this: 2000-01-01 Av1 2000-02-01 Bv1 2000-03-01 Av2 2000-04-01 Cv1 2000-05-01 Av3 2000-06-01 Bv2 2000-07-01 Bv3 2000-08-01 Cv2 So start with the newest. Now back through the timeline .. so the sets to check, in order, would be: Av3,Bv3,Cv2 Av3,Bv3,Cv1 Av3,Bv2,Cv1 Av3,Bv1,Cv1 Av2,Bv1,Cv1 Note that A1 never gets tested, cause A2 was out already by the time B & C existed. Now, maybe A1,B1,C1 will work, too .. but is that something we'd be concerned with? What's the common view (if any) on providing min version numbers for the sake of not blindly saying "any"? my assumption was that the min version was usually when the author knows he's using a feature that's non-existent or broken previous to a certain version. Using the min version from backpan that works may not be the true min version ... But also, may not care about those edge cases.	[reply]

Replies are listed 'Best First'.

Re: Find minimum required module version
by davidrw (Prior) on Dec 08, 2008 at 17:43 UTC

cpan-testers

What do you test again? So your module Z requires A, B, C. Do you prep w/the newest of B & C, and then try all versions of A (newest->oldest), running the test suite Z until it breaks?
This would definitely require pretty decent coverage by Z's test suite.
Might there be recursive problems with this? Maybe Z works fine with any B ... but A requires (for functionality that Z doesn't use; but still a requirement for install) a certain version of B. How does that get handled?
Maybe such a script, instead of finding the min versions for each prereq independently, finds a point in time that they work?
Construct a time line of release dates like this:
```
2000-01-01  Av1
2000-02-01  Bv1
2000-03-01  Av2
2000-04-01  Cv1
2000-05-01  Av3
2000-06-01  Bv2
2000-07-01  Bv3
2000-08-01  Cv2
```
So start with the newest. Now back through the timeline .. so the sets to check, in order, would be:
- Av3,Bv3,Cv2
- Av3,Bv3,Cv1
- Av3,Bv2,Cv1
- Av3,Bv1,Cv1
- Av2,Bv1,Cv1
Note that A1 never gets tested, cause A2 was out already by the time B & C existed. Now, maybe A1,B1,C1 will work, too .. but is that something we'd be concerned with?
What's the common view (if any) on providing min version numbers for the sake of not blindly saying "any"?
my assumption was that the min version was usually when the author knows he's using a feature that's non-existent or broken previous to a certain version.
Using the min version from backpan that works may not be the true min version ...
But also, may not care about those edge cases.

[reply]

Back to Seekers of Perl Wisdom