Drop scala shadowing and DottyPredef #10428
Conversation
odersky
force-pushed
the
drop-scalaShadowing
branch
3 times, most recently
from
37f1203 to
3b46af1
Compare
| def withFinalizer(finalizer: Finalizer): SymbolLoader = | ||
| val cpy = clone().asInstanceOf[SymbolLoader] | ||
| cpy.finalizer = { sym => | ||
| this.finalizer(sym) |
There was a problem hiding this comment.
should there be a local reference to this.finalizer to avoid capturing the whole SymbolLoader if unnecessary?
odersky
force-pushed
the
drop-scalaShadowing
branch
from
1cab33b to
623ddd2
Compare
|
@abgruszecki Can you review this with an eye whether some doc generation needs to be adapted? |
| @@ -1199,6 +1199,7 @@ class Namer { typer: Typer => | |||
| cls.setNoInitsFlags(parentsKind(parents), untpd.bodyKind(rest)) | |||
| if (cls.isNoInitsClass) cls.primaryConstructor.setFlag(StableRealizable) | |||
| processExports(using localCtx) | |||
| defn.patchStdLibClass(denot.asClass) | |||
There was a problem hiding this comment.
What happens when we pickle Predef.scala to Tasty, then unpickle it? As far as I can tell, since we're only adding symbols and not trees, the pickled Predef won't have the extra definitions, so we need to run patchStdlibClass in TreeUnpickler too.
There was a problem hiding this comment.
But do we ever unpickle Predef from Tasty? Why would we want to do that?
There was a problem hiding this comment.
The doctool should probably see the unpickled Predef, for definition lookup purposes. I'd also worry about any user of Tasty Inspector that wants to look up definitions in Predef.
There was a problem hiding this comment.
But so far nobody pickles Predef into Tasty, right?
There was a problem hiding this comment.
I checked and it seems a Predef.tasty file is created, we also document it in the doctool already:
https://scala3doc.virtuslab.com/pr-master/scala3-stdlib/api/scala/-predef/index.html?query=%20object%20Predef%20extends%20LowPriorityImplicits
There was a problem hiding this comment.
The PR looks good to me, aside from the remark that Guillaume already had. If we can ensure in this PR that definitions loaded from Tasty will have the same patches applied, I'd say we should do that. Otherwise, I can try doing that myself when doing more work on definition lookup in the doctool.
odersky
force-pushed
the
drop-scalaShadowing
branch
from
7c35990 to
60b5ac6
Compare
odersky
force-pushed
the
drop-scalaShadowing
branch
from
e26fd1e to
77ecfea
Compare
|
I will merge this as soon as the CI is green. I have been spending about 50% of my time yesterday and today to re-apply the same patc with thishes endlessly to the CB only to see them being rendered outdated by other changes. |
It's been replaced a while ago by `-source 3.0-migration`
Once we have a new bootstrap compiler we can drop it entirely.
Instead of using DottyPredef, patch the Predef class when unpickling it.
The patch adds all methods from a patch class
scala.runtime.stdLibPatches.Predef
to scala.Predef.
Instead, patch the `scala.language` object in the same way we patch `scala.Predef`. We leave a dummy file in scalaShadowing, since the build requires the package to exist. When we have a new bootstrap compiler, we should drop `scalaShadowing` entirely.
Patch its members recursively instead.
With so much churn it's almost impossible to keep up.
odersky
force-pushed
the
drop-scalaShadowing
branch
from
18b375f to
c351094
Compare
Several steps:
1. Make bootstrapped DottyPredef an empty object
Once we have a new bootstrap compiler we can drop it entirely.
Instead of using DottyPredef, patch the Predef class when unpickling it.
The patch adds all methods from a patch class
to scala.Predef.
2. Drop scalaShadowing after bootstrap
Instead, patch the
scala.languageobject in the same way we patchscala.Predef.We leave a dummy file in scalaShadowing, since the build requires the package to exist.
When we have a new bootstrap compiler, we should drop
scalaShadowingentirely.