You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As discussed in #384, there's a number of things to consider about transitioning glide onto gps. I figured I'd open this issue as a place to start just enumerating them, and we can decide where/how to split them all out later.
I intend to edit and update this list as I go, though, so probably better to think of this as a meta-TODO list, where we can check things off as we talk through them and agree on something, implement something, or whatever other solution we come to.
(This list is unlikely to be exhaustive until everything's actually checked, as it's a running checklist that I'll add to as we discover more things)
Cannot specify 'vcs' param; this is deduced, or doesn't work. At most, it becomes a deducer question
subpkgs (in manifest) no longer serve a purpose and are going away. They may remain in the lock, though. gps doesn't currently support that (Express set of packages actually used in result/lock sdboyer/gps#40), but adding support is pretty trivial (we have all the information).
gps makes hard assumptions about things being available upstream, else fail. (So, if something's available in vendor, but not accessible upstream, it's a failure, whereas arguably it could/should be satisfied by what's sitting in vendor now).
glide.yaml key is changed: s/Import/Dependencies/. It might be even more semantically accurate if it were DependencyConstraints or Constraints
It seems like glide's current ignore and exclude in glide.yaml could both just be folded into gps' ignore, but I'm not fully sure.
This affects glide get, b/c it's kinda a no-op if it's not in the import graph - apart from updating the manifest. As described in gps issue, though, I think we could maybe safely defer that till later?
In the same vein, if a dependency is listed in glide.yaml, but is not present in the import graph, it will NOT show up in the result.
User-facing output will be significantly different, most notably in that gps doesn't report that it's doing any repo fetching or management as it does it. We could build a communication channel for it if it's crucial, but it just doesn't exist now.
As discussed in #384, there's a number of things to consider about transitioning glide onto gps. I figured I'd open this issue as a place to start just enumerating them, and we can decide where/how to split them all out later.
I intend to edit and update this list as I go, though, so probably better to think of this as a meta-TODO list, where we can check things off as we talk through them and agree on something, implement something, or whatever other solution we come to.
(This list is unlikely to be exhaustive until everything's actually checked, as it's a running checklist that I'll add to as we discover more things)
gps doesn't currently support that (Express set of packages actually used in result/lock sdboyer/gps#40), but adding support is pretty trivial (we have all the information).glide.yamlkey is changed: s/Import/Dependencies/. It might be even more semantically accurate if it wereDependencyConstraintsorConstraintsignoreandexcludeinglide.yamlcould both just be folded into gps' ignore, but I'm not fully sure.glide get, b/c it's kinda a no-op if it's not in the import graph - apart from updating the manifest. As described in gps issue, though, I think we could maybe safely defer that till later?glide.yaml, but is not present in the import graph, it will NOT show up in the result.