#general
- e
enough-analyst-54434
07/23/2015, 1:18 AMYeah - that's the single bug I'm aware of - e
enough-analyst-54434
07/23/2015, 1:21 AMI hadn't thought about it but I think intransitive is another ivy superset thing that does not map - afaict there is no intransitive in poms, you must instead exclude all direct direct deps of the dep manually - e
enough-analyst-54434
07/23/2015, 1:23 AMSo if we're gunning to only model pom, we may have no ivy bugs in-play - e
enough-analyst-54434
07/23/2015, 1:24 AM... and the last big one I can think of is force - superset - e
enough-analyst-54434
07/23/2015, 1:25 AMBasically 3 widely used things at Twitter are all part of the superset and probably lead to confusion many times because of that - they don't do pom-like dep management things - e
enough-analyst-54434
07/23/2015, 1:26 AMMade more sense when the internal repo was an ivy repo in 2010/2011 - b
bored-art-40741
07/23/2015, 2:05 AMThere is a notion of intransitive in poms - b
bored-art-40741
07/23/2015, 2:05 AMexcluding ('*', '*') is considered equivalent to being intransitive - e
enough-analyst-54434
07/23/2015, 2:20 AMAha - did not know wildcards worked - e
enough-analyst-54434
07/23/2015, 2:20 AMThis of course has a direct equivalent in ivy using excludes and the regexp matcher - e
enough-analyst-54434
07/23/2015, 2:22 AMaha - 2013 - this is recentish: https://issues.apache.org/jira/browse/MNG-2315 - b
bored-art-40741
07/23/2015, 2:22 AMtbh the pom ecosystem is so broken that it's difficult to imagine a tool that isn't just layers of hacks - b
bored-art-40741
07/23/2015, 2:23 AMAny strict interpretation of the pom standard will quickly find many transitive dependencies on maven central that have broken POMs - e
enough-analyst-54434
07/23/2015, 2:23 AMs/pom/every transitive dep system I've seen where anyone can add to the global dep graph/ - b
bored-art-40741
07/23/2015, 2:23 AMThat is true - e
enough-analyst-54434
07/23/2015, 2:23 AMautomatic transitive dep systems are a lie - b
bored-art-40741
07/23/2015, 2:24 AMI guess what I'm getting at is that the more general your resolver gets, the more impossible it is for it to exist - b
bored-art-40741
07/23/2015, 2:24 AMIn my experience, reducing scope is the only way to get a sane resolver - b
bored-art-40741
07/23/2015, 2:24 AMWhich is to say, finding the balance between manual configuration and automatic resolution - b
bored-art-40741
07/23/2015, 2:25 AMAs much as our custom pom resolver is lacking features I'd like (but don't have the time to add), it has been an overwhelming success in terms of engineer hours spent on classpath issues - e
enough-analyst-54434
07/23/2015, 2:25 AMNo doubt - b
bored-art-40741
07/23/2015, 2:26 AMAny change that results in a change to the global classpath must include, in the commit message, the diff between the classpaths (which the tool will generate for you). Which is unbelievably helpful when issues crop up and the question becomes, "is it classpath related?" - e
enough-analyst-54434
07/23/2015, 2:28 AMYup - e
enough-analyst-54434
07/23/2015, 2:30 AMGoogle had this except no tools, you just manually add the new checked-in dep sourcecode, download the source jar and check it in basically, see if CI goes green - b
bored-art-40741
07/23/2015, 2:32 AMyeah, the longer I've worked on a large codebase, the more I've appreciated the power of the monorepo. It's not great for open source culture, but the end goal for everyone is to be more productive, and sometimes that means not staying glued to upstream unless upstream has significant generality, momentum, and quality release process - w
witty-crayon-22786
07/23/2015, 2:33 AMwe're hoping to revisit sapling soon. - b
bored-art-40741
07/23/2015, 2:34 AMYeah, I was thinking of that - w
witty-crayon-22786
07/23/2015, 2:34 AMjohn dropped a bunch of notes which he probably wouldn't mind me sharing - b
bored-art-40741
07/23/2015, 2:34 AMsapling was pretty cool, though it's doesn't fix the fundamental issue of only testing against your internal repo - e
enough-analyst-54434
07/23/2015, 2:34 AMYeah - I'd still love to solve the problem of putting the company 1st but treating open source as 1st class as possible