StackOverflowError when using Java code that creates types with LazyRef #21521
Comments
wjoel
added
itype:bug
itype:crash
stat:needs triage
Gedochao
added
area:typer
and removed
stat:needs triage
|
Myself and @hamzaremmal were not able to reproduce the issue. In any of the 4 configurations with either Scala 3.5.0 or the latest nightly, and with either Java 17 or Java 22; we get the following error: -- [E008] Not Found Error: /Users/eugeneflesselle/Documents/dotty/tests/playground/example.scala:11:5
7 | Box.builder()
8 | .setOrientation(Orientation.VERTICAL)
9 | .setHalign(Align.START)
10 | .setValign(Align.BASELINE)
11 | .se
| ^
|value se is not a member of org.gnome.gtk.Box#Builder[? <: org.gnome.gtk.Box#Builder[?]]#B#B².
|Extension methods were tried, but the search failed with:
|
| Recursion limit exceeded.
| Maybe there is an illegal cyclic reference?
| If that's not the case, you could also try to increase the stacksize using the -Xss JVM option.
| For the unprocessed stack trace, compile with -Xno-decode-stacktraces.
| A recurring operation is (inner to outer):
|
| collectParts of org.gnome.gtk.Widget#Builder[
| org.gnome.gtk.Box#Builder[? <: org.gnome.gtk.Box#Builder[?]]#B#B]
| collectParts of <:
| org.gnome.gtk.Widget#Builder[
| org.gnome.gtk.Box#Builder[? <: org.gnome.gtk.Box#Builder[?]]#B#B]
|
|where: B is a type in trait Builder² with bounds <: io.github.jwharm.javagi.gobject.Builder³[B]
| B² is a type in class Builder⁴ with bounds <: org.gnome.gtk.Widget#Builder⁴[B²]The stack trace in the original issue points to code using |
|
Thanks for looking into this, @EugeneFlesselle! That is odd. It shouldn't matter, but I'm on Linux, Fedora 40. And with this in #!/bin/bash
mkdir i21521
cd i21521
echo '//> using scala 3.5.0
//> using dep io.github.jwharm.javagi:gtk:0.10.2
import org.gnome.gtk.{Unit as _, *}
object Main {
Box.builder()
.setOrientation(Orientation.VERTICAL)
.setHalign(Align.START)
.setValign(Align.BASELINE)
.se
.build()
.asInstanceOf[Box]
}' > main.scala && scala-cli compile .I get: Restarting bloop (it wasn't actually started when I ran this the first time, but just in case, I tried again after That being said, there might be a (false) cyclic reference hidden in the LazyRefs. I say false, because it works with other versions of Scala 3, and it works in Java. Does my PR take care of your supposedly cyclic reference and stack overflow being reported? |
|
@wjoel Ah! I was able to reproduce it, thank you for the additional details. Could you try it with bloop disabled completely
It does, I'll add a review over on the PR directly. |
|
I still get an overflow, it's just printed differently (no suggestion to restart the bloop server, for example, and with more details): |
|
With |
|
I've bisected this, and the bad commit is 9d88c80, "Drop redundant ... I get the same error message as on 3.4.2. So I'm guessing that |
This reverts commit 9d88c80. Closes #21521 The `ClassTypeParamCreationFlags` include both `TypeParam` and `Deferred`. In effect, a class type parameter was considered an anchor for implicit search, by `sym.is(Deferred)` as a sufficient condition. For a failing example, one can try asserting: ```scala || sym.is(Deferred).ensuring(_ == sym.is(Deferred, butNot = Param)) ``` in `ImplicitRunInfo#isAnchor` and a test with `summon[Ordering[Int]]`. In that example, at least, the flags happen to be set by `Scala2Unpickler#readDisambiguatedSymbol`: https://github.com/scala/scala3/blob/614170f4545ea6da8f07e0c4b0f2fdfe01377270/compiler/src/dotty/tools/dotc/core/unpickleScala2/Scala2Unpickler.scala#L560
Compiler version
3.5.0, works on 3.4.2, 3.4.3.
Minimized code
I have unfortunately not been able to minimize this to a self contained example.
Note that
.se(which does not exist, and should be a compiler error) is intentional, and triggers the stack overflow.Scala 3.4.2 fails with a reasonable error message:
Output (click arrow to expand)
The text was updated successfully, but these errors were encountered: