Extensible snapshots #3249
This is a significant change touching large parts of the code base and relating to many issues. I'm going to do a write-up of what ended up getting covered here (I'll link to it from this issue when ready), as well as try to collect issues that I believe are related.
At the time of opening this PR, all tests and integration tests pass on my OS X machine, and I'm currently using this branch for my main development. Nonetheless, there are still some outstanding issues, such as a Store hash mismatch on GHC 7.10, a few FIXMEs scattered through the PR that I'd like to improve upon. Also, additional real world testing is needed.
There's not really any kind of logical grouping of these commits that I could rebase to. The best I could do is split off the first two commits if desired and squash the rest down into one mega-commit. If people prefer that, let me know.
EDIT Here's a work in progress write-up about the new feature: https://fpcomplete-site-extensible-snapshots.review.gitlab.fpcomplete.com/blog/2017/07/stacks-new-extensible-snapshots
The text was updated successfully, but these errors were encountered:
This came in automatically at the inception of Stack, by copying data types from stackage-curator. In reality, we don't need all of this information within Stack. As we move towards extensible snapshots, we want to make sure we have the bare minimum of information to allow a shared data type for parsing all kinds of snapshots. Additionally, we probably want to move away from depending on extra information present in the build plan, in case there are mistakes in it.
Because (1) we don't necessarily trust it (may be wrong upstream, or using a different GHC install locally), and (2) custom snapshots don't want to provide this. This patch is not complete on its own. In particular, it's possible (such as with the script command) to try to load up the global package information before GHC is installed. Next step is to have a proper separation between a resolved and unresolved build plan.