- name: Checkout Code
         uses: actions/checkout@v7
+        with:
+          persist-credentials: false

Apply to both the canonical and comparison job checkout steps.

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

🏁 Script executed:

#!/bin/bash
set -euo pipefail

echo "## git status"
git status --short

echo
echo "## repo files of interest"
git ls-files | rg '^(AGENTS\.md|docs/coding-patterns\.md|asm/buffer\.go|asm/.*\.go|.*memory.*\.go)$'

echo
echo "## outline asm/buffer.go"
ast-grep outline asm/buffer.go --view expanded || true

echo
echo "## outline candidate memory files"
for f in $(git ls-files | rg '(^|/)memory.*\.go$|(^|/)mmap.*\.go$|(^|/)buffer.*\.go$'); do
  echo "--- $f"
  ast-grep outline "$f" --view expanded || true
done

echo
echo "## search for free methods in asm"
rg -n "func \\(.*\\) free\\(" asm || true

echo
echo "## search for Buffer.Free usage"
rg -n "func \\(b \\*Buffer\\) Free\\(|\\.Free\\(\\)" asm || true

Repository: siyul-park/minivm

Length of output: 5450

🏁 Script executed:

#!/bin/bash
set -euo pipefail

echo "## docs/coding-patterns relevant excerpt"
if [ -f docs/coding-patterns.md ]; then
  rg -n -A3 -B3 "Free|resource|cleanup|error|mmap|buffer" docs/coding-patterns.md || true
fi

echo
echo "## asm/buffer.go relevant section"
sed -n '1,180p' asm/buffer.go

echo
echo "## files containing type Buffer or mmap wrapper"
rg -n "type Buffer|type .*mmap|func \\(.*\\) free\\(" asm || true

Repository: siyul-park/minivm

Length of output: 8969

🏁 Script executed:

#!/bin/bash
set -euo pipefail

echo "## search for double-free guards / unmap implementation"
rg -n "munmap|Unmap|syscall\\.Munmap|Close\\(|freed|already.*free|idempot" . || true

Repository: siyul-park/minivm

Length of output: 22507

🏁 Script executed:

#!/bin/bash
set -euo pipefail

echo "## targeted grep for memory.free implementation"
rg -n -A20 -B10 "func \\(.*\\) free\\(" asm || true

Repository: siyul-park/minivm

Length of output: 3856

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

🏁 Script executed:

#!/bin/bash
# Inspect reject()'s definition and the outer Compile() driver's use of c.ip
# to confirm whether constant()'s KindAny branch needs an explicit c.ip advance.
rg -n "func reject" internal/cmd/geninterp/*.go
rg -n "func .*Compile\(" internal/cmd/geninterp/*.go asm/*.go interp/*.go
rg -n "\.ip \+=|\.ip =" internal/cmd/geninterp/lower.go | head -60

Repository: siyul-park/minivm

Length of output: 626

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Inspect the relevant code paths around reject() and Compile().
sed -n '1848,1898p' internal/cmd/geninterp/lower.go
sed -n '220,340p' interp/jit.go
sed -n '340,430p' interp/jit.go

Repository: siyul-park/minivm

Length of output: 8431

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Inspect the surrounding lowering flow and all places where c.ip is advanced.
sed -n '300,470p' internal/cmd/geninterp/lower.go
printf '\n---\n'
rg -n "c\\.ip \\+=|ip \\+=|width\\)" internal/cmd/geninterp/lower.go

Repository: siyul-park/minivm

Length of output: 7939

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Inspect the lowering epilogue around the shared c.ip increment and how result.compile is emitted.
sed -n '1760,1835p' internal/cmd/geninterp/lower.go
printf '\n---\n'
sed -n '1835,1885p' internal/cmd/geninterp/lower.go

Repository: siyul-park/minivm

Length of output: 4325

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Inspect the rest of finish() and nearby emitted ip-advance logic.
sed -n '470,540p' internal/cmd/geninterp/lower.go
printf '\n---\n'
sed -n '540,620p' internal/cmd/geninterp/lower.go

Repository: siyul-park/minivm

Length of output: 5252

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Map the benchmark file structure first.
ast-grep outline benchmarks/compare_test.go --view expanded || true

# Show the relevant sections with line numbers.
sed -n '1,260p' benchmarks/compare_test.go | nl -ba | sed -n '1,260p'

# Find the exact helper/closure names involved.
rg -n "int32\\(|lua\\.LNumber|ToInteger\\(|py\\.Int|goja\\.Undefined|result :=|call\\(" benchmarks/compare_test.go

Repository: siyul-park/minivm

Length of output: 1718

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Probe the semantics of the involved conversions in a read-only way.
cat > /tmp/verify_numeric_conversions.go <<'EOF'
package main

import (
	"fmt"
	"math"
)

func main() {
	fmt.Println("int32(42.5) =", int32(42.5))
	fmt.Println("int32(2147483648) =", int32(2147483648))
	fmt.Println("int32(-2147483649) =", int32(-2147483649))
	fmt.Println("math.Trunc(42.5) == 42.5:", math.Trunc(42.5) == 42.5)
}
EOF

go run /tmp/verify_numeric_conversions.go

Repository: siyul-park/minivm

Length of output: 437

🏁 Script executed:

#!/bin/bash
set -euo pipefail
sed -n '110,230p' benchmarks/compare_test.go | nl -ba

Repository: siyul-park/minivm

Length of output: 195

🏁 Script executed:

#!/bin/bash
set -euo pipefail
python3 - <<'PY'
print("probe")
PY

Repository: siyul-park/minivm

Length of output: 161

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Inspect the exact benchmark comparison sites.
sed -n '120,225p' benchmarks/compare_test.go | awk '{printf "%4d  %s\n", NR+119, $0}'

Repository: siyul-park/minivm

Length of output: 3149

🏁 Script executed:

#!/bin/bash
set -euo pipefail

sed -n '120,225p' benchmarks/compare_test.go | awk '{printf "%4d  %s\n", NR+119, $0}'

Repository: siyul-park/minivm

Length of output: 3149

🌐 Web query:

gpython py.Int underlying type documentation

💡 Result:

In the gpython project, which is a partial implementation of the Python 3.4 interpreter written in Go, py.Int serves as a wrapper type to represent Python integers [1][2][3]. Underlyingly, py.Int is defined as a type based on the Go int64 primitive [1]. As a Go-based implementation of Python, gpython uses native Go types for several Python objects to facilitate performance and integration [1]. Specifically, the gpython codebase utilizes py.Int as an immutable type that is passed by value within the interpreter's runtime [1]. While py.Int corresponds to standard integer values, gpython also includes py.BigInt for cases requiring arbitrary-precision integers, and internal operations often convert between these types or to an int64 representation during hash calculations [1]. For further technical context, you can examine the gpython source code repository directly, specifically files within the py/ directory where such core types are defined [1][2][4].

Citations:

• 1: https://github.com/go-python/gpython/blob/main/notes.txt
• 2: https://github.com/go-python/gpython/
• 3: https://pkg.go.dev/github.com/go-python/gpython
• 4: https://github.com/go-python/gpython/tree/main/examples/embedding

🌐 Web query:

goja Value.ToInteger truncates float documentation

💡 Result:

In the Goja library, the Value.ToInteger method behaves by converting a value to an integer by truncating the fractional part of a floating-point number, consistent with standard ECMAScript ToInteger conversion rules [1][2]. When called on a floating-point value, Goja internally utilizes logic that effectively truncates the number towards zero [1][2]. Specifically, implementations typically use a helper function (often named floatToIntClip or similar) that converts the float64 to an int64 [1][2]. This process: 1. Handles special cases: NaN values are converted to 0 [1][2]. 2. Clamps the result: Values exceeding the range of int64 are clamped to math.MaxInt64 or math.MinInt64 [1][2]. 3. Truncates: For finite numbers within the integer range, it performs a standard truncation of the decimal component [1][2]. This behavior aligns with the JavaScript specification, which requires that conversion to an integer type discards the fractional part [3]. Because Goja is designed to execute ECMAScript, it adheres to these specifications [3]. Developers using Goja should be aware that this truncation to an integer can result in a loss of precision, and in some contexts (such as setting timeouts or intervals in integrated systems like k6), passing non-integer values may lead to unexpected results or the need for explicit float-based handling if fractional precision is required [3].

Citations:

• 1: https://github.com/packing/goja/blob/d3859c04cace/value.go
• 2: https://github.com/powerpuffpenguin/goja/blob/491e284ada3c/value.go
• 3: Fix ws.Set(Timeout/Interval) panicking on float values grafana/k6#1608

🌐 Web query:

goja Value.ToInteger method docs ECMAScript truncates NaN Infinity

💡 Result:

In the Goja library, the Value.ToInteger method does not simply truncate values like NaN or Infinity; instead, it adheres to specific handling logic designed to convert ECMAScript values to a 64-bit integer (int64) [1][2][3]. When ToInteger is called on a floating-point value in Goja, it utilizes an internal function—commonly referred to in the source code as floatToIntClip—which processes the number using the following rules [3][4]: 1. NaN: If the value is NaN, it returns 0 [3][4]. 2. Infinity: If the value is positive infinity, it is clipped to the maximum value of a 64-bit integer (math.MaxInt64). If it is negative infinity, it is clipped to the minimum value of a 64-bit integer (math.MinInt64) [3][4]. 3. Finite Numbers: If the number is finite, it is cast to int64, while also being clipped to the bounds of the int64 range (math.MinInt64 to math.MaxInt64) [3][4]. This behavior differs from a simple truncation or a raw float-to-int conversion, as it explicitly handles non-finite numbers by clipping them to the edges of the representable 64-bit integer range [3][4]. You can find these implementations in the value.go file within the Goja source code [2][3][4].

Citations:

• 1: https://pkg.go.dev/github.com/dop251/goja
• 2: https://github.com/packing/goja/blob/d3859c04cace/value.go
• 3: https://github.com/packing/goja/blob/8ce18d86df5f3bbc6ad925fcb61771aa4baf1c8c/value.go
• 4: https://github.com/elastic/goja/blob/241b3421/value.go

🌐 Web query:

pkg.go.dev goja Value ToInteger int64 truncates

💡 Result:

The ToInteger method on the goja Value interface returns an int64 representation of the JavaScript value, adhering to ECMAScript specifications which involve truncation towards zero [1][2]. In the Goja implementation, ToInteger performs the conversion by truncating any fractional part, effectively rounding towards zero [1][3]. For example: - Integer values (valueInt) are returned as their underlying int64 representation [1][2]. - Floating-point values (valueFloat) are converted by truncating the decimal portion [1][2]. - Non-numeric types are converted according to JavaScript rules (e.g., null, undefined, or false often result in 0) [1][2]. Because the return type of ToInteger is int64, any JavaScript number that exceeds the range of an int64 (or is NaN/Infinity) will be subjected to the limitations of the Go int64 type after the conversion process, which typically involves standard Go integer overflow behavior if the resulting truncated value cannot be represented [1][2]. In practical usage within Goja, ToInteger is intended to implement the ECMAScript ToInteger abstract operation, which explicitly specifies truncation of fractional parts [1][3].

Citations:

• 1: https://github.com/packing/goja/blob/c1e8210f876286fa5227e473b5cc2b5bb85b4af2/value.go
• 2: https://github.com/packing/goja/blob/d3859c04cace/value.go
• 3: https://deepwiki.com/elastic/goja/2.2-value-system

🌐 Web query:

gopher-lua LNumber underlying float64 docs

💡 Result:

In GopherLua, the LNumber type is explicitly defined as a float64 [1][2][3]. This design choice mirrors the behavior of the original Lua 5.1 specification, where numbers are represented as double-precision floating-point values [4]. Because LNumber is a float64, it can represent integers exactly only up to the limit of safe integer precision for floating-point numbers (typically 2^53) [5][4]. Developers should be aware that conversions involving large integers may result in a loss of precision [4]. Key implementation details include: - Underlying Type: LNumber is a type alias for float64 [1][2]. - Type System: It satisfies the LValue interface [1][2]. - Utility Functions: The library provides helper functions like isInteger to check if an LNumber can be represented as an integer without losing fractional data [5][6]. - String Formatting: When converting to a string, the library checks if the value is an integer; if so, it formats it as an integer, otherwise it formats it as a float [7][8].

Citations:

• 1: https://pkg.go.dev/github.com/yuin/gopher-lua@v1.0.0
• 2: https://github.com/yuin/Gopher-lua
• 3: https://github.com/yuin/gopher-lua/tree/v1.1.2
• 4: int64 last digit changes (treated as float64) yuin/gopher-lua#174
• 5: https://github.com/yuin/gopher-lua/blob/master/utils.go
• 6: https://github.com/yuin/gopher-lua/blob/v1.1.2/utils.go
• 7: https://github.com/yuin/gopher-lua/blob/master/value.go
• 8: https://github.com/yuin/gopher-lua/blob/v1.1.2/value.go

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Show the benchmark cases and the expressions that determine expected outputs.
rg -n "benchmark(GopherLua|Goja|Gpython|Yaegi|Tengo|Native|Wazero)|want|fib|sieve|sum|branchTree|iterative|recursive|typedArray" benchmarks/compare_test.go

# Print the whole file in manageable chunks around the benchmark definitions.
sed -n '1,120p' benchmarks/compare_test.go
sed -n '120,240p' benchmarks/compare_test.go
sed -n '240,420p' benchmarks/compare_test.go
sed -n '420,620p' benchmarks/compare_test.go

Repository: siyul-park/minivm

Length of output: 19537

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Inspect benchmark sources used by the comparison helpers.
sed -n '1,240p' benchmarks/compare_test.go | awk '{printf "%4d  %s\n", NR, $0}'

Repository: siyul-park/minivm

Length of output: 7182

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Collect exact lines for the benchmark cases and return-value handling.
python3 - <<'PY'
from pathlib import Path
p = Path('benchmarks/compare_test.go')
for i, line in enumerate(p.read_text().splitlines(), 1):
    if any(k in line for k in ('benchmarkGopherLua', 'benchmarkGoja', 'benchmarkGpython', 'benchmarkYaegi',
                               'benchmarkTengo', 'benchmarkNative', 'benchmarkWazero',
                               'want', 'fib', 'sieve', 'sum', 'branchTree', 'iterative', 'recursive', 'typedArray')):
        print(f"{i:4d}  {line}")
PY

Repository: siyul-park/minivm

Length of output: 3265

🏁 Script executed:

#!/bin/bash
set -euo pipefail

python3 - <<'PY'
from pathlib import Path
p = Path('benchmarks/compare_test.go')
for i, line in enumerate(p.read_text().splitlines(), 1):
    if any(k in line for k in (
        'benchmarkGopherLua', 'benchmarkGoja', 'benchmarkGpython', 'benchmarkYaegi',
        'benchmarkTengo', 'benchmarkNative', 'benchmarkWazero',
        'want', 'fib', 'sieve', 'sum', 'branchTree', 'iterative', 'recursive', 'typedArray'
    )):
        print(f"{i:4d}  {line}")
PY

Repository: siyul-park/minivm

Length of output: 3265

• benchmarks/compare_test.go#L128-L132 (this comment)
• benchmarks/compare_test.go#L167-L172
• benchmarks/compare_test.go#L204-L205
• benchmarks/compare_test.go#L215-L216