fix: if no deps, avoid npe; also make sure POMs refd exist #1901
Conversation
cstamas
changed the title
cstamas
force-pushed
the
foolproof-deps
branch
from
4ee6af9 to
465a134
Compare
cstamas
changed the title
|
The strict resolving seems to break some tests, right? |
Like versionless artifact was sent to load POM for, while this now fails (before it was just silent about it due forgiving policy)
|
Even if now it works, this is wrong. The deps with ranges should be first "materialized" as exact versions, as nothing guarantees that you load POM of v1 and then resolve v2 (this is very bent example, but still, no guarantee for it). Or in other words, now the range is resolved twice: once to get POM (to get depMgt from it) and once during collection. |
This format is really really weird...
|
sorry for being daft here - whats the wrong part? afaik we didn't do anything different with version ranges than normal maven...but you write as if we did - so curius to grok it before merging? |
|
no problem per se, feel free to merge. It is only me tinkering... All in all, is fine to merge, but needs more. Will add PR later. |
|
@cstamas it seems these changes have caused issues with some of our users. Could you perhaps take a look at this issue and see what's going on? enola-dev/enola#1040 |
|
I don't think so. In fact, i think these changes made some issues in projects apparent, that were hidden before... We will see... |
|
So I was right 😄 |
|
Awesome! :-) I don't understand though how it worked in JBang version 0.122.0? Does that have to do with the "lax resolver" PR you opened? If so I would love if you could explain the "how" to me, because it's a mystery to me 😅 |
If there are no deps (or only imports, which does not make sense), do not belly up. Also, make sure that all the refd POMs (of deps and imports) are present, as by def Maven uses "forgiving" policy.