test(harness): port check_generators to Perl

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

3 files changed, 89 insertions, 28 deletions

diff --git a/tests/CMakeLists.txt b/tests/CMakeLists.txt
index 630d363..1cb2eff 100644
--- a/tests/CMakeLists.txt
+++ b/tests/CMakeLists.txt
@@ -5,8 +5,8 @@ set(CPSC_TEST_COUNT 10000 CACHE STRING "Random records per differential group")
 
 add_test(NAME harness COMMAND "${Python3_EXECUTABLE}"
     "${CMAKE_CURRENT_SOURCE_DIR}/test_runner.py")
-add_test(NAME generators COMMAND "${Python3_EXECUTABLE}"
-    "${CMAKE_CURRENT_SOURCE_DIR}/check_generators.py" "${PERL_EXECUTABLE}")
+add_test(NAME generators COMMAND "${PERL_EXECUTABLE}"
+    "${CMAKE_CURRENT_SOURCE_DIR}/check_generators.pl" "${PERL_EXECUTABLE}")
 foreach(suite i ii iii)
     file(GLOB cases CONFIGURE_DEPENDS "${CMAKE_CURRENT_SOURCE_DIR}/${suite}/*/case.json")
     foreach(manifest IN LISTS cases)
diff --git a/tests/check_generators.pl b/tests/check_generators.pl
new file mode 100644
index 0000000..731215b
--- /dev/null
+++ b/tests/check_generators.pl
@@ -0,0 +1,87 @@
+#!/usr/bin/env perl
+# Require identical generated input bytes across platforms and Perl versions.
+use strict;
+use warnings;
+use Digest::SHA qw(sha256_hex);
+use File::Spec;
+use File::Temp ();
+use FindBin;
+use JSON::PP ();
+
+die "usage: $0 <perl>\n" unless @ARGV == 1;
+my ($perl) = @ARGV;
+
+my $dir = File::Spec->catdir($FindBin::Bin, 'ii');
+my $vectors_path = File::Spec->catfile($dir, 'generator-vectors.json');
+open my $fh, '<:raw', $vectors_path or die "cannot read $vectors_path: $!\n";
+my $vectors_text = do { local $/; <$fh> };
+close $fh;
+my $vectors = JSON::PP->new->decode($vectors_text);
+
+for my $name (sort keys %{$vectors}) {
+    my $digest = $vectors->{$name};
+    my $generate = File::Spec->catfile($dir, $name, 'generate.pl');
+    my @outputs;
+    for my $seed (('0' x 40, '0' x 40, '1' x 40)) {
+        my ($exit, $out, $err) = run_capture($perl, $generate, $seed, '100');
+        die "$name: generator exited with $exit\n" if $exit != 0;
+        if (length $err) {
+            die "$name: unexpected stderr " . display($err) . "\n";
+        }
+        push @outputs, $out;
+    }
+    if ($outputs[0] ne $outputs[1] || $outputs[0] eq $outputs[2]) {
+        die "$name: seed reproducibility failed\n";
+    }
+    if (sha256_hex($outputs[0]) ne $digest) {
+        die "$name: generated input differs from the portable test vector\n";
+    }
+    print "PASS $name: repeatable, seed-sensitive, portable bytes\n";
+}
+
+sub run_capture {
+    my (@command) = @_;
+    my $out_fh = File::Temp->new(UNLINK => 1);
+    my $err_fh = File::Temp->new(UNLINK => 1);
+    my $out_path = $out_fh->filename;
+    my $err_path = $err_fh->filename;
+    close $out_fh;
+    close $err_fh;
+    my $pid = fork();
+    die "cannot fork: $!\n" unless defined $pid;
+    if ($pid == 0) {
+        open STDOUT, '>', $out_path or exit 127;
+        open STDERR, '>', $err_path or exit 127;
+        binmode STDOUT;
+        binmode STDERR;
+        exec { $command[0] } @command;
+        exit 127;
+    }
+    my $timed_out = 0;
+    local $SIG{ALRM} = sub { $timed_out = 1; kill 'KILL', $pid; };
+    alarm 10;
+    waitpid($pid, 0);
+    alarm 0;
+    if ($timed_out) {
+        waitpid($pid, 0);
+        die "generator timed out after 10s: @command\n";
+    }
+    my $status = $?;
+    open my $out_in, '<:raw', $out_path or die "cannot read $out_path: $!\n";
+    my $out = do { local $/; <$out_in> };
+    close $out_in;
+    open my $err_in, '<:raw', $err_path or die "cannot read $err_path: $!\n";
+    my $err = do { local $/; <$err_in> };
+    close $err_in;
+    $out = '' unless defined $out;
+    $err = '' unless defined $err;
+    my $exit = ($status >> 8);
+    return ($exit, $out, $err);
+}
+
+sub display {
+    my ($bytes) = @_;
+    $bytes =~ s/([^ -~\n])/sprintf('\\x{%02x}', ord($1))/ge;
+    $bytes =~ s/\n/\\n/g;
+    return "'$bytes'";
+}
diff --git a/tests/check_generators.py b/tests/check_generators.py
deleted file mode 100644
index 2b92789..0000000
--- a/tests/check_generators.py
+++ /dev/null
@@ -1,26 +0,0 @@
-#!/usr/bin/env python3
-"""Require identical generated input bytes across platforms and Perl versions."""
-
-import hashlib
-import json
-from pathlib import Path
-import subprocess
-import sys
-
-
-directory = Path(__file__).resolve().parent / "ii"
-vectors = json.loads((directory / "generator-vectors.json").read_text(encoding="utf-8"))
-for name, digest in vectors.items():
-    command = [sys.argv[1], str(directory / name / "generate.pl")]
-    outputs = []
-    for seed in ["0" * 40, "0" * 40, "1" * 40]:
-        result = subprocess.run([*command, seed, "100"], check=True, stdout=subprocess.PIPE,
-                                stderr=subprocess.PIPE, timeout=10)
-        if result.stderr:
-            raise RuntimeError(f"{name}: unexpected stderr {result.stderr!r}")
-        outputs.append(result.stdout)
-    if outputs[0] != outputs[1] or outputs[0] == outputs[2]:
-        raise RuntimeError(f"{name}: seed reproducibility failed")
-    if hashlib.sha256(outputs[0]).hexdigest() != digest:
-        raise RuntimeError(f"{name}: generated input differs from the portable test vector")
-    print(f"PASS {name}: repeatable, seed-sensitive, portable bytes")