PackageManager.getBestPackage: Introduce Version[Range] overloads #2375
Conversation
source/dub/packagemanager.d
Outdated
| @@ -224,6 +224,7 @@ class PackageManager { | |||
|
|
|||
| /** Looks up the first package matching the given name. | |||
| */ | |||
| deprecated("The semantic of this function are unhelpful, prefer `getBestPackage` with `name, Dependency.any` instead") | |||
There was a problem hiding this comment.
| deprecated("The semantic of this function are unhelpful, prefer `getBestPackage` with `name, Dependency.any` instead") | |
| deprecated("The semantic of this function are unhelpful, prefer `getBestPackage` with `name, VersionRange.any` instead") |
source/dub/dub.d
Outdated
| logDiagnostic("External sub package %s %s not found.", name, dep.version_); | ||
| auto ret = dep.visit!( | ||
| (VersionRange rng) => m_dub.m_packageManager.getBestPackage(name, rng), | ||
| (any) => assert(0, "Cannot call getPackageRaw with a non-version dependency on subpackages"), |
There was a problem hiding this comment.
it looks like it'd be possible that this assert(false) branch is called from
getSpecificConfigs -> getPackage -> getPackageRaw
There was a problem hiding this comment.
It shouldn't be, because otherwise the assert(0) in Dependency.version_ would get triggered.
But I'll look at the CI failure.
There was a problem hiding this comment.
the dependency resolution logic is tricky to understand, so I don't really know where to start searching for such a case, but in general these non-version cases are not that well tested, so an edge case is unlikely to trigger in CI here
Geod24
force-pushed
the
easy-fix-getBestPackage
branch
from
d679b1e to
4888a1e
Compare
Geod24
force-pushed
the
easy-fix-getBestPackage
branch
2 times, most recently
from
76be451 to
2071c5b
Compare
Geod24
force-pushed
the
easy-fix-getBestPackage
branch
from
2071c5b to
346a2d3
Compare
Geod24
changed the title
getBestPackage and deprecate getFirstPackage
|
@WebFreak001 : Removed deprecation, this should be a no-brainer now (and fix the "TODO"). |
|
I plan to deprecate it, but the previous iteration didn't pass the CI for unknown reason, so I'll do the deprecation later (it was a side product, not the main target, of the refactoring). |
Continuing on my crusade to have correctly-typed interfaces when it comes to dependencies.