From b98494db2332e1f7877d098914fc4f4a01a9f66c Mon Sep 17 00:00:00 2001 From: Earlopain <14981592+Earlopain@users.noreply.github.com> Date: Thu, 12 Feb 2026 16:57:14 +0100 Subject: [PATCH 1/2] Run tests against some bundled gems They make extensive use of prism. --- .github/workflows/cruby-bindings.yml | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/.github/workflows/cruby-bindings.yml b/.github/workflows/cruby-bindings.yml index d635e550f0..e9cd55d8f6 100644 --- a/.github/workflows/cruby-bindings.yml +++ b/.github/workflows/cruby-bindings.yml @@ -57,3 +57,7 @@ jobs: EXCLUDES: ${{ matrix.parser == 'parse.y' && './test/.excludes-parsey' }} run: make -j$(nproc) -s test-all working-directory: ruby/ruby + - name: make test-bundled-gems + if: ${{ matrix.parser == 'prism' }} + run: make -j$(nproc) -s test-bundled-gems BUNDLED_GEMS=irb,rbs,rdoc,syntax_suggest + working-directory: ruby/ruby From 82d953fcad3dd64fb8cc53f34f0b02b245b573c5 Mon Sep 17 00:00:00 2001 From: Earlopain <14981592+Earlopain@users.noreply.github.com> Date: Wed, 11 Feb 2026 13:03:49 +0100 Subject: [PATCH 2/2] Fix lexing for unterminated strings/heredocs etc. When we hit EOF and still have lex modes left, it means some content was unterminated. Heredocs specifically have logic that needs to happen when the body finished lexing. If we don't reset the mode back to how it was before, it will not continue lexing at the correct place. --- src/prism.c | 9 ++++ .../unterminated_heredoc_and_embexpr.txt | 11 ++++ .../unterminated_heredoc_and_embexpr_2.txt | 9 ++++ test/prism/lex_test.rb | 53 +++++++++++++++++-- 4 files changed, 79 insertions(+), 3 deletions(-) create mode 100644 test/prism/errors/unterminated_heredoc_and_embexpr.txt create mode 100644 test/prism/errors/unterminated_heredoc_and_embexpr_2.txt diff --git a/src/prism.c b/src/prism.c index 34e5d38b0a..097b4c7305 100644 --- a/src/prism.c +++ b/src/prism.c @@ -9856,6 +9856,15 @@ parser_lex(pm_parser_t *parser) { // We'll check if we're at the end of the file. If we are, then we // need to return the EOF token. if (parser->current.end >= parser->end) { + // We may be missing closing tokens. We should pop modes one by one + // to do the appropriate cleanup like moving next_start for heredocs. + // Only when no mode is remaining will we actually emit the EOF token. + if (parser->lex_modes.current->mode != PM_LEX_DEFAULT) { + lex_mode_pop(parser); + parser_lex(parser); + return; + } + // If we hit EOF, but the EOF came immediately after a newline, // set the start of the token to the newline. This way any EOF // errors will be reported as happening on that line rather than diff --git a/test/prism/errors/unterminated_heredoc_and_embexpr.txt b/test/prism/errors/unterminated_heredoc_and_embexpr.txt new file mode 100644 index 0000000000..bed7fcd24e --- /dev/null +++ b/test/prism/errors/unterminated_heredoc_and_embexpr.txt @@ -0,0 +1,11 @@ +<= "3.3" - def test_lex_compare - prism = Prism.lex_compat(File.read(__FILE__), version: "current").value - ripper = Ripper.lex(File.read(__FILE__)) + def test_lex_compat + source = "foo bar" + prism = Prism.lex_compat(source, version: "current").value + ripper = Ripper.lex(source) assert_equal(ripper, prism) end end + + def test_lex_interpolation_unterminated + assert_equal( + %i[STRING_BEGIN EMBEXPR_BEGIN EOF], + token_types('"#{') + ) + + assert_equal( + %i[STRING_BEGIN EMBEXPR_BEGIN IGNORED_NEWLINE EOF], + token_types('"#{' + "\n") + ) + end + + def test_lex_interpolation_unterminated_with_content + # FIXME: Emits EOL twice. + assert_equal( + %i[STRING_BEGIN EMBEXPR_BEGIN CONSTANT EOF EOF], + token_types('"#{C') + ) + + assert_equal( + %i[STRING_BEGIN EMBEXPR_BEGIN CONSTANT NEWLINE EOF], + token_types('"#{C' + "\n") + ) + end + + def test_lex_heredoc_unterminated + code = <<~'RUBY'.strip + <