fix(#18265): crash on extension method without type nor RHS #18743
Conversation
This commit fixes a bug that causes Scala compiler to crash due to position error when an extension method followed by a newline has neither type annotation nor `= <expr>`. This commit is based on scala#18445. Co-Authored-By: @hamzaremmal
| @@ -2577,6 +2577,8 @@ object Parsers { | |||
| parents match { | |||
| case parent :: Nil if !in.isNestedStart => | |||
| reposition(if (parent.isType) ensureApplied(wrapNew(parent)) else parent) | |||
| case tkn if in.token == INDENT => | |||
There was a problem hiding this comment.
With brace syntax, the last token consumed by a parser is }, but with indent syntax, the last token consumed by a parser is OUTDENT, which causes mismatching spans（actual offset is n, but reposition set offset less than n）. Template returned from templateBodyOpt has the exact span, so we reuse it.
| } = /*<-local6*/new: | ||
| def scalameta/*<-local5*/ = "4.0" | ||
| } = new: | ||
| /*<-local6*/def scalameta/*<-local5*/ = "4.0" |
There was a problem hiding this comment.
I just run scala3-compiler/Test/runMain dotty.tools.dotc.semanticdb.updateExpect according to https://dotty.epfl.ch/docs/contributing/testing.html#semanticdb-tests-1. To be honest, I'm not sure what this change means.
Expect files are used as regression tests to detect changes in the compiler.
Their correctness is determined by human inspection
Which is correct, the current one or the previous one...?:thinking:
There was a problem hiding this comment.
If <-local6 is supposed to point to a local anonymous class, I feel the current one is correct.
i10416
changed the title
close #18265
This commit fixes a bug that causes Scala compiler to crash due to position error when an extension method followed by a newline has neither type annotation nor
= <expr>.This commit is based on #18445.
Co-Authored-By: @hamzaremmal