Make more parts of stdlib compile #1826

odersky · 2016-12-18T17:27:32Z

No description provided.

odersky · 2016-12-19T10:29:02Z

The continuous-integration failure is because the drone does not pull updated versions of the scala-scala stdlib.

odersky · 2016-12-19T14:45:21Z

@felixmulder There's a dottyDoc failure, which causes the junit test to fail:

[error] Test dotty.tools.dottydoc.TestWhitelistedCollections.arrayHasDocumentation failed: java.lang.AssertionError: assertion failed: invalid prefix PolyType(List(1), List(T), List(TypeBounds(TypeRef(ThisType(TypeRef(NoPrefix,scala)),Nothing), TypeRef(ThisType(TypeRef(NoPrefix,scala)),Any))), RefinedType(RefinedType(TypeRef(ThisType(TypeRef(NoPrefix,scala)),PartialFunction), scala$PartialFunction$$A, TypeAlias(TypeRef(TermRef(ThisType(TypeRef(NoPrefix,scala)),package)/withSig(Signature(List(),)),Throwable), -1)), scala$PartialFunction$$B, TypeAlias(PolyParam(T), 1))), took 8.569 sec
[error]     at scala.Predef$.assert(Predef.scala:165)
[error]     at dotty.tools.dotc.core.Types$NamedType.<init>(Types.scala:1386)
[error]     at dotty.tools.dotc.core.Types$TypeRef.<init>(Types.scala:1759)

Also, we need to pull in the latest scala-scala so that local CI succeeds.

smarter · 2016-12-19T14:57:53Z

Could the CI run git pull on scala-scala first?

felixmulder · 2016-12-19T14:59:03Z

It could - but better to have a version local to the image and then update it. I'll add that soonish

…

On Mon, Dec 19, 2016, 15:57 Guillaume Martres ***@***.***> wrote: Could the CI run git pull on scala-scala first? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1826 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABdYwUXL_L2tgSdXVa7e6oPeF2MFgJjQks5rJptzgaJpZM4LQLlc> .

odersky · 2016-12-19T16:06:46Z

Rebased to master. Also, blacklisted SystemProperties again which apparently caused the dottydoc failure.

felixmulder · 2016-12-19T17:29:39Z

The CI can now update scala-scala (needs latest drone.yml from master)

felixmulder · 2016-12-19T17:30:04Z

Regarding the doc-tool failure - I'll have a look at it tomorrow morning and add the commit to this PR

felixmulder · 2016-12-19T20:20:51Z

I tried enabling the SystemProperties and compiling it with dottydoc - doesn't seem to crash the compiler anymore. But there is a stack overflow when compiling std lib from dottydoc. Not sure if it is relevant to the PR currently. Let me know if there's anything else I can do to help out here.

odersky · 2016-12-20T08:24:02Z

@felixmulder On validate-junit it still says:

[error] Test dotty.tools.dottydoc.TestWhitelistedCollections.arrayHasDocumentation failed: java.lang.AssertionError: assertion failed: invalid prefix PolyType(List(1), List(T), List(TypeBounds(TypeRef(ThisType(TypeRef(NoPrefix,scala)),Nothing), TypeRef(ThisType(TypeRef(NoPrefix,scala)),Any))), RefinedType(RefinedType(TypeRef(ThisType(TypeRef(NoPrefix,scala)),PartialFunction), scala$PartialFunction$$A, TypeAlias(TypeRef(TermRef(ThisType(TypeRef(NoPrefix,scala)),package)/withSig(Signature(List(),)),Throwable), -1)), scala$PartialFunction$$B, TypeAlias(PolyParam(T), 1))), took 15.693 sec
[error] at scala.Predef$.assert(Predef.scala:165)

Same for continuous-integration/drone/pr. I don't see a stackoverfloa anywhere (?).

But it might well be unrelated to sys/SystemProperties.

felixmulder · 2016-12-20T08:52:24Z

Hmm, not sure - the CI (both of them) give different results then when I run it locally on my machine. On my machine I get three stack overflows. One when compiling the standard library (compileStdLib) and then twice again when running the dottydoc tests - which also compile the whitelist, so that seemed reasonable. On the CI it is instead as you say, some invalid prefix. This seems to be the offending line:

[error]     at dotty.tools.dottydoc.core.DocASTPhase.dotty$tools$dottydoc$core$DocASTPhase$$membersFromSymbol$1(DocASTPhase.scala:49)

Which is this line: https://github.com/lampepfl/dotty/blob/master/doc-tool/src/dotty/tools/dottydoc/core/DocASTPhase.scala#L49

I'm going to try it again with a clean repo.

UPDATE: on a clean build I get the same error as the CIs

odersky · 2016-12-20T16:14:07Z

🎉 Thanks for the quick fixes @felixmulder !

odersky · 2016-12-20T16:18:12Z

Current status: We have some bootstrap related issues, and at least two unrelated transform problems (one in tailrec, the other probably in elimByName). And then there are some files that compile only if there's no deep subtype check, we have to figure whether this is normal or a bug.

But I think the current state is ready to go in. Who wants to give it a shot for reviewing?

Under language:Scala2, don't require an explicit `override' when overriding default methods of Java traits. `scalac` does currently the same thing.

When faced with a denotation that combines parameterless and nullary method definitions (toString is a common example), ignore any redundant () applications.

We looked under the wrong signature before, which meant that we sometimes would not detect that a super accessor existed already and generate a duplicate. Observed when compiling stdlib's collection/mutable/ObservableMap.scala.

If the expected type is Unit, any parameterless member should be considered applicable when doing overloading resolution.

Avoids potentially expensive string assembly operations.

1. Overriding a Java-8 concrete method now produces a migration warning under -language:Scala2, 2. Overriding a normal with a lazy val is now a migration warning instead of an error.

- Move non-specialized functions to whitelist - Replace `macro ???` with just `???` in scala-scala Several other consolidations

Argument comparison of hk types did not take into account that the compared types could have themselves wildcard arguments.

SystemProperties leads to a dottydoc failure, so stays blacklisted for now. It does pass the normal compile-stdlib test.

smarter · 2016-12-20T22:09:08Z

I'll review tomorrow

Fixing two failures in stdlib having to do with widening. In both cases, a widenIfUnstable was required. One was widening too much, the other too little.

Need to be careful not to read a classfile before a compilation unit defining the annotation is entered.

odersky · 2016-12-21T12:02:13Z

@smarter I did not want to do that because it's ugly and inefficient, and, in 99.99% of cases, not needed.

odersky · 2016-12-21T14:27:24Z

I could not find out what the problems with whitelisting Function1 and Function2 are. the junit test passes but partest fails with a cyclic reference error. Can somebody with more partest fu figure out where that comes from?

Anyway, if we would compile Function1/2 as part of a bootstrap it would be done sequentially, so I don't see much of an issue.

That means the only real oustanding issue are the three tailcall failures. It would be good to adress those.

odersky · 2016-12-21T14:41:00Z

/rebuild

odersky · 2016-12-21T14:41:52Z

We get a project loading failed on continuous integration. @felixmulder any way to do the equivalent of /rebuild here?

smarter · 2016-12-21T14:43:26Z

Felix tells me that our CI is running out of disk space, so out of order for now

smarter · 2016-12-21T14:45:37Z

Fabien is on it.

felixmulder · 2016-12-21T14:54:37Z

Fabien tells us it should be fixed now. Looking into it further. Restarting the jobs is easy using either the drone CLI tool or from the web interface. Cheers, Felix

…

On Wed, Dec 21, 2016 at 3:45 PM Guillaume Martres ***@***.***> wrote: Fabien is on it. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1826 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABdYwS6OR4kQJRg7Hc766PgDGQPhzLIHks5rKTuTgaJpZM4LQLlc> .

smarter

Otherwise LGTM

smarter · 2016-12-21T18:09:39Z

compiler/src/dotty/tools/dotc/typer/Typer.scala

+   *  0th strategy: If `tree` overrides a nullary method, mark the prototype
+   *                so that the argument is dropped and return `tree` itself.
+   *
+   *  After that, two strategies are tried, and the firs that is successful is picked.


typo: firs -> first

smarter · 2016-12-21T18:23:50Z

compiler/src/dotty/tools/dotc/typer/FrontEnd.scala

@@ -75,7 +84,11 @@ class FrontEnd extends Phase {
    }
    unitContexts foreach (parse(_))
    record("parsedTrees", ast.Trees.ntrees)
-    unitContexts.foreach(enterSyms(_))


What's wrong with foreach here?

We need to keep track of remaining.

odersky added the stat:in progress label Dec 18, 2016

odersky force-pushed the fix-compile-stdlib branch from 03f3091 to 308547f Compare December 19, 2016 16:06

odersky added stat:needs review and removed stat:in progress labels Dec 20, 2016

odersky added 14 commits December 20, 2016 18:33

Don't require override for Java interface default methods

c386699

Under language:Scala2, don't require an explicit `override' when overriding default methods of Java traits. `scalac` does currently the same thing.

Blacklist scala/util/Sorting.scala (reverted from commit b960b12)

11d8ed5

Fix typo in file name

9c0df5a

More lenient handling of mixed parameterless and nullary methods

7c5e36b

When faced with a denotation that combines parameterless and nullary method definitions (toString is a common example), ignore any redundant () applications.

Fix super accessor generation

d9f66a2

We looked under the wrong signature before, which meant that we sometimes would not detect that a super accessor existed already and generate a duplicate. Observed when compiling stdlib's collection/mutable/ObservableMap.scala.

Treat Unit type specially in overloading resolution

c4da13f

If the expected type is Unit, any parameterless member should be considered applicable when doing overloading resolution.

Update scala-collections lists

1f5c542

More reshuffling of scala-collections lists

318965e

Make msg in testScala2Mode by-name.

8e2a8db

Avoids potentially expensive string assembly operations.

Add two refcheck migration warnings

b2e4cf1

1. Overriding a Java-8 concrete method now produces a migration warning under -language:Scala2, 2. Overriding a normal with a lazy val is now a migration warning instead of an error.

More reorgs to scala-collectons lists

1fec473

- Move non-specialized functions to whitelist - Replace `macro ???` with just `???` in scala-scala Several other consolidations

Fix subtyping of hk types with wildcard arguments

dbdb123

Argument comparison of hk types did not take into account that the compared types could have themselves wildcard arguments.

Further massaging of stdlib-collections lists

f42a568

Blacklist SystemProperties again

a086383

SystemProperties leads to a dottydoc failure, so stays blacklisted for now. It does pass the normal compile-stdlib test.

odersky and others added 2 commits December 20, 2016 18:33

Reorganize blacklist along failure classes

a8457a3

Fix dottydoc ast phase

190ee9d

felixmulder force-pushed the fix-compile-stdlib branch from 3428580 to 190ee9d Compare December 20, 2016 17:33

smarter self-assigned this Dec 20, 2016

odersky added 2 commits December 21, 2016 09:25

Fiddling with widenening

cf4f773

Fixing two failures in stdlib having to do with widening. In both cases, a widenIfUnstable was required. One was widening too much, the other too little.

Tweak the way annotations are represented in desugaring

ec35b84

Need to be careful not to read a classfile before a compilation unit defining the annotation is entered.

odersky force-pushed the fix-compile-stdlib branch from c80a754 to ec35b84 Compare December 21, 2016 14:24

Drop redundant println

40af028

smarter approved these changes Dec 21, 2016

View reviewed changes

Address reviewer comment

9568eba

smarter merged commit 404ce76 into scala:master Dec 21, 2016

smarter removed the stat:needs review label Dec 21, 2016

allanrenucci deleted the fix-compile-stdlib branch December 14, 2017 19:19

Make more parts of stdlib compile #1826

Make more parts of stdlib compile #1826

Uh oh!

Conversation

odersky commented Dec 18, 2016

Uh oh!

odersky commented Dec 19, 2016

Uh oh!

odersky commented Dec 19, 2016 • edited by felixmulder Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

smarter commented Dec 19, 2016

Uh oh!

felixmulder commented Dec 19, 2016 via email

Uh oh!

odersky commented Dec 19, 2016

Uh oh!

felixmulder commented Dec 19, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

felixmulder commented Dec 19, 2016

Uh oh!

felixmulder commented Dec 19, 2016

Uh oh!

odersky commented Dec 20, 2016

Uh oh!

felixmulder commented Dec 20, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

odersky commented Dec 20, 2016

Uh oh!

odersky commented Dec 20, 2016

Uh oh!

smarter commented Dec 20, 2016

Uh oh!

odersky commented Dec 21, 2016

Uh oh!

odersky commented Dec 21, 2016

Uh oh!

odersky commented Dec 21, 2016

Uh oh!

odersky commented Dec 21, 2016

Uh oh!

smarter commented Dec 21, 2016

Uh oh!

smarter commented Dec 21, 2016

Uh oh!

felixmulder commented Dec 21, 2016 via email

Uh oh!

smarter left a comment

Choose a reason for hiding this comment

Uh oh!

smarter Dec 21, 2016

Choose a reason for hiding this comment

Uh oh!

smarter Dec 21, 2016

Choose a reason for hiding this comment

Uh oh!

odersky Dec 21, 2016

Choose a reason for hiding this comment

Uh oh!

Uh oh!

odersky commented Dec 19, 2016 •

edited by felixmulder

Loading

felixmulder commented Dec 19, 2016 •

edited

Loading

felixmulder commented Dec 20, 2016 •

edited

Loading