You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
On a Windows console, console.log and every other line the CLIs write now show characters outside the console's code page, such as 😀 or 日本語, instead of ?. When GetConsoleMode succeeds on standard output or error, the text goes out as UTF-16 through WriteConsoleW. The code page is never changed.
Redirected output stays UTF-8 bytes. OutputCallback is unchanged, and Linux and macOS don't link the new unit.
Where: the new ConsoleOutput unit replaces the write function of a standard text file whose handle is a console. The RTL still fills the buffer with UTF-8, and each flush hands its complete characters to a writer that converts them on the stack, so nothing is allocated, and calls WriteConsoleW. ProcessEncoding applies it to Output, StdOut, ErrOutput and StdErr.
Why not in EmitLine: every Write and WriteLn goes through the text file, so timing lines, test reports, errors and the REPL are covered without touching a call site.
Out of scope: the RTL reopens the standard files for each new thread, so other threads keep the console's code page. No thread in the repository writes to them directly. GUI applications are left alone.
Not checked on a real Windows console. Under wine 9.0 in a pseudo-terminal, ü, 日本語 and ✓ print where main prints ?. Wine shows any surrogate pair from WriteConsoleW as two U+FFFD, so the emoji is unconfirmed.
Visual walkthrough
script.js:
console.log("ü 😀 日本語");
GocciaRunner script.js in cmd (code page 850)
GocciaRunner script.js > out.txt
main
ü, and ? for each character missing from code page 850
UTF-8 bytes of ü 😀 日本語
This PR
ü 😀 日本語 (subject to the console font)
the same UTF-8 bytes
How to test
On Windows, in cmd or PowerShell with the default OEM code page, run build\GocciaRunner.exe script.js with the script above. Expected: ü 😀 日本語, and chcp afterwards prints the same code page as before.
Run build\GocciaRunner.exe script.js > out.txt. Expected: out.txt holds the UTF-8 bytes C3 BC 20 F0 9F 98 80 20 E6 97 A5 E6 9C AC E8 AA 9E 0D 0A.
On any platform: ./build.pas tests --test=ConsoleOutput.Test --run-tests. Expected: all tests pass.
Edit: ac1e358 reviewed
Code review (fresh engineering-quality lane plus coordinator): nine findings fixed, none open. Fixed: a held-back byte dropped at close, a heap allocation per console write, GUI standard files switched to UTF-8, and test gaps.
Full CI (ci.yml, Windows x64 and i386-win32 included): pending for this head.
… page on a Windows console
When a standard handle is a console, ConsoleOutput replaces the text
file's write function: the RTL fills the buffer with UTF-8 and each
flush writes the complete characters as UTF-16 with WriteConsoleW. Every
Write and WriteLn to the console goes through it, including the CLIs'
timing lines and reports. Redirected output stays UTF-8 bytes, and the
console's code page is never changed.
Closes#1890
…thing per write, leave GUI standard files alone
The console write function holds back a split character only when the RTL
flushes a full buffer, so Close and the flush at exit write everything. The
Windows writer converts UTF-8 to UTF-16 on the stack. ProcessEncoding only
changes the standard files of a console application. The test starts its
file in code page 850, decodes each writer call on its own, and names the
IOResult of a failed write.
frostney
added
bug
Something isn't working
os specific
This issue is only for a specific platform/OS
runtime
TGocciaRuntime: optional host extensions (console, fetch, JSON5, FFI, test assertions, etc.)
labels
Nov 5, 2026
No actionable comments were generated in the recent review.
Walkthrough
Windows console output now converts UTF-8 text to UTF-16 and writes it with WriteConsoleW. Redirected output remains UTF-8. Initialization configures standard output streams, and documentation describes the encoding behavior.
Changes: Windows console output
Layer / File(s)
Summary
UTF-8 output buffering and validation source/shared/ConsoleOutput.pas, source/shared/ConsoleOutput.Test.pas
A new helper configures UTF-8 output and handles incomplete trailing UTF-8 sequences when buffers split characters. Tests cover console and redirected output, flushing, close-time tails, and writer failures.
Windows stream configuration source/shared/ProcessEncoding.pas, docs/contributing/tooling.md, docs/embedding.md
Windows console output is converted to UTF-16 and written through WriteConsoleW. Initialization configures four output streams for console processes. The documentation describes console, redirected, and GUI standard-file behavior.
No actionable merge-blocking issue remains for the repository’s current output paths. A caller that installs an unusually small console text buffer could still stall a write.
Architecture risk: Low · up to ac1e3
The change affects 2 systems.
Changed systems:source, docs
Architecture concerns
No architecture-level concerns identified.
Review details
Systems and components
observed — source (service) was modified; 3 changed files map to changed impact.
observed — docs (service) was modified; 2 changed files map to changed impact.
Before / after behavior
observed — Modified behavior in docs/contributing/tooling.md: The Windows encoding description now applies UTF-8 to RTL-returned names and, in console applications, to all four standard text files; redirected output is UTF-8, while GUI standard files retain RTL settings. Console handles use ConsoleOutput to convert buffered UTF-8 to UTF-16 via WriteConsoleW without changing the console code page. The text also adds the per-thread reopening detail and identifies Output and StdOut among the configured files.
observed — Modified behavior in docs/embedding.md: The documentation adds that Windows console output uses UTF-16 through WriteConsoleW, and that linking the engine sets the RTL code page to UTF-8 and configures the main thread’s standard output and error; previously it mentioned only UTF-8 output when redirected.
observed — Modified behavior in source/shared/ConsoleOutput.Test.pas: Adds the test program’s imports, console code-page and text fixtures, test-suite declarations, and shared probe/writer state.
observed — Modified behavior in source/shared/ConsoleOutput.Test.pas: Adds probe functions that record the handle and distinguish console from redirected output, a writer that captures bytes and decodes each call separately, and helpers for UTF-8 encoding, raw byte construction, and complete-length checks.
Pre-merge checks: 5
Check name
Status
Explanation
Title check
The title clearly and concisely describes the main Windows console output fix.
Description check
The description includes the required Summary section, explains the change, and documents key implementation constraints and non-goals.
Linked Issues check
Issue #1890 requires Unicode text on interactive Windows stdout and stderr, unchanged console code page, and UTF-8 bytes for redirected output. ProcessEncoding.pas installs UseUnicodeOutput for `O…
Out of Scope Changes check
The changes stay within issue #1890. The new ConsoleOutput unit, Windows integration, tests, and documentation directly implement or validate console Unicode output and redirected UTF-8 behavior. Th…
Docstring Coverage
No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentat…
Full details: Linked Issues check
Explanation
Issue #1890 requires Unicode text on interactive Windows stdout and stderr, unchanged console code page, and UTF-8 bytes for redirected output. ProcessEncoding.pas installs UseUnicodeOutput for Output, StdOut, ErrOutput, and StdErr in console applications. ConsoleOutput.pas detects console handles and routes complete UTF-8 sequences to WriteConsoleW; non-console handles keep UTF-8 text-file output. The tests cover console detection, redirected bytes, split UTF-8 characters, line flushing, close handling, and writer failure. The implementation covers other Write and WriteLn calls through the standard text files. The documentation records the thread and GUI limitations.
Full details: Out of Scope Changes check
Explanation
The changes stay within issue #1890. The new ConsoleOutput unit, Windows integration, tests, and documentation directly implement or validate console Unicode output and redirected UTF-8 behavior. The embedding documentation describes the changed output behavior. No unrelated product behavior or unrelated files are changed.
Full details: Docstring Coverage
Explanation
No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 0 functions across 0 files. (5 skipped: 5 unsupported.)
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Reviewing files that changed from the base of the PR and between 7c6c822 and ac1e358.
Files selected for processing: 5
docs/contributing/tooling.md
docs/embedding.md
source/shared/ConsoleOutput.Test.pas
source/shared/ConsoleOutput.pas
source/shared/ProcessEncoding.pas
Included review availability: This review used your included allowance. 0 included reviews remain after this review. Your included PR review attempts over the past 7 days set your current allowance at 1 review per hour.
Steady-state failures and timeouts are non-blocking; PASS → non-timeout failure transitions fail the conformance gate. Measured on ubuntu-latest x64. Areas grouped by the first two test262 path components; minimum 25 attempted tests, areas already at 100% excluded. Δ vs main compares against the most recent cached main baseline.
18 pinned Web Tooling workloads; 18 workloads produced at least one Goccia sample. Raw results from 1 sample per workload; full stdout/stderr for failures and min/max/CV stay in the web-tooling-report artifact. This lane is Goccia-only by design.
14 pinned AWFY benchmarks. Medians from 5 interleaved samples per engine; raw JSON includes min/max/CV and is attached as the awfy-report artifact. Δ compares PR Goccia median against the same-runner main Goccia median; overlapping min/max ranges are treated as unchanged noise. Hardware counters were requested but not recorded: perf stat failed: Error: No supported events found. Rows below 0.5ms are timer-floor sensitive.
1.00× means aligned; values above 1.00× mean Goccia was proportionally slower after normalizing JetStream’s higher-is-better score. This is a directional barometer across runtimes with different goals, not a product ranking. Raw samples and failure details remain in the jetstream-report artifact.
Deterministic profile diff: no significant changes.
Measured on ubuntu-latest x64. Each PR run also builds the main base and benchmarks it back-to-back on the same runner after a warm-up discard, so the ranges compare two runs measured under the same conditions; overlapping min/max ranges are treated as unchanged noise. Percentage deltas are secondary context. See docs/adr/0076.
GC rows aggregate the main thread plus all worker thread-local GCs. Test runner worker shutdown frees thread-local heaps in bulk; that shutdown reclamation is not counted as GC collections or collected objects.
Metric
Value
GC Live
23.8 KiB
GC Peak Live
522.51 MiB
GC Allocated During Run
3.22 GiB
GC Limit
7.80 GiB
GC Collections
3,671
GC Collected Objects
17,729,882
Heap Start Allocated
369.4 KiB
Heap End Allocated
3.04 MiB
Heap Delta Allocated
2.68 MiB
Heap Delta Free
2.14 MiB
Benchmarks (447)
Metric
Value
Total
447
Workers
4
Duration
2.87min
Memory
GC rows aggregate the main thread plus all worker thread-local GCs. Benchmark runner performs explicit between-file collections, so collection and collected-object counts can be much higher than the test runner.
Metric
Value
GC Live
5.6 KiB
GC Peak Live
230.73 MiB
GC Allocated During Run
41.08 GiB
GC Limit
7.80 GiB
GC Collections
10,691
GC Collected Objects
575,857,414
Heap Start Allocated
1.22 MiB
Heap End Allocated
1.22 MiB
Heap Delta Allocated
3.3 KiB
Boot
Metric
Value
Boot (median)
3.9ms
Boot (min)
3.9ms
Empty-script (0;) end-to-end process wall-clock for GocciaRunner, min and median over 60 runs after 10 warmups. Includes a fixed process / dynamic-loader floor that does not shrink, so it understates the engine-boot speedup; treat as practical boot time, not pure engine work.
Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.
Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
bugSomething isn't workingos specificThis issue is only for a specific platform/OSruntimeTGocciaRuntime: optional host extensions (console, fetch, JSON5, FFI, test assertions, etc.)
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
frostney
added
bug
Summary
console.logand every other line the CLIs write now show characters outside the console's code page, such as😀or日本語, instead of?. WhenGetConsoleModesucceeds on standard output or error, the text goes out as UTF-16 throughWriteConsoleW. The code page is never changed.OutputCallbackis unchanged, and Linux and macOS don't link the new unit.Closes #1890
Details
ConsoleOutputunit replaces the write function of a standard text file whose handle is a console. The RTL still fills the buffer with UTF-8, and each flush hands its complete characters to a writer that converts them on the stack, so nothing is allocated, and callsWriteConsoleW.ProcessEncodingapplies it toOutput,StdOut,ErrOutputandStdErr.EmitLine: everyWriteandWriteLngoes through the text file, so timing lines, test reports, errors and the REPL are covered without touching a call site.ü,日本語and✓print wheremainprints?. Wine shows any surrogate pair fromWriteConsoleWas twoU+FFFD, so the emoji is unconfirmed.Visual walkthrough
script.js:GocciaRunner script.jsincmd(code page 850)GocciaRunner script.js > out.txtmainü, and?for each character missing from code page 850ü 😀 日本語ü 😀 日本語(subject to the console font)How to test
cmdor PowerShell with the default OEM code page, runbuild\GocciaRunner.exe script.jswith the script above. Expected:ü 😀 日本語, andchcpafterwards prints the same code page as before.build\GocciaRunner.exe script.js > out.txt. Expected:out.txtholds the UTF-8 bytesC3 BC 20 F0 9F 98 80 20 E6 97 A5 E6 9C AC E8 AA 9E 0D 0A../build.pas tests --test=ConsoleOutput.Test --run-tests. Expected: all tests pass.Edit:
ac1e358reviewedci.yml, Windows x64 and i386-win32 included): pending for this head.