feat(driver): embed and automatically link the standard library

Add -no-std to disable default library loading and linking.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

12 files changed, 119 insertions, 4 deletions

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 5d9f84f..365e6b1 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -144,6 +144,16 @@ foreach(source IN LISTS cpsc_library_sources)
 endforeach()
 add_custom_target(pseudocode-lib ALL DEPENDS ${cpsc_library_outputs})
 
+set(cpsc_embedded_library "${CMAKE_BINARY_DIR}/standard-library/StandardLibrary.inc")
+add_custom_command(OUTPUT "${cpsc_embedded_library}"
+    COMMAND "${CMAKE_COMMAND}" -E make_directory "${CMAKE_BINARY_DIR}/standard-library"
+    COMMAND "${PERL_EXECUTABLE}" "${CMAKE_SOURCE_DIR}/tools/driver/embed-standard-library.pl"
+        "${cpsc_embedded_library}" ${cpsc_library_outputs}
+    DEPENDS ${cpsc_library_outputs} tools/driver/embed-standard-library.pl
+    VERBATIM)
+target_sources(cpsc PRIVATE tools/driver/StandardLibrary.cc "${cpsc_embedded_library}")
+target_include_directories(cpsc PRIVATE "${CMAKE_BINARY_DIR}/standard-library")
+
 option(BUILD_TESTING "Enable compiler conformance tests" OFF)
 if(BUILD_TESTING)
     enable_testing()
diff --git a/pseudocode-lib/std.cps b/pseudocode-lib/std.cps
new file mode 100644
index 0000000..955e7da
--- /dev/null
+++ b/pseudocode-lib/std.cps
@@ -0,0 +1,2 @@
+#LIB
+INCLUDE "_exp_string.cps"
diff --git a/tests/i/no-std/case.json b/tests/i/no-std/case.json
new file mode 100644
index 0000000..85acec5
--- /dev/null
+++ b/tests/i/no-std/case.json
@@ -0,0 +1 @@
+{"compiler_args":["-no-std"]}
diff --git a/tests/i/no-std/expected.txt b/tests/i/no-std/expected.txt
new file mode 100644
index 0000000..1a2735d
--- /dev/null
+++ b/tests/i/no-std/expected.txt
@@ -0,0 +1,2 @@
+C literal without the standard library
+42
diff --git a/tests/i/no-std/notes.md b/tests/i/no-std/notes.md
new file mode 100644
index 0000000..8fcea00
--- /dev/null
+++ b/tests/i/no-std/notes.md
@@ -0,0 +1,3 @@
+# No standard library
+
+Human review: pending. A suffixed literal is C_STRING without loading any pseudocode library. The C puts call writes the supplied text and a newline; integer addition produces 42.
diff --git a/tests/i/no-std/program.cps b/tests/i/no-std/program.cps
new file mode 100644
index 0000000..feb369f
--- /dev/null
+++ b/tests/i/no-std/program.cps
@@ -0,0 +1,4 @@
+INCLUDE_OBJ "__libc"
+EXTERN "C" EXTERN_FUNCTION puts(C_STRING) RETURNS C_INT
+CALL puts("C literal without the standard library"C)
+OUTPUT 20 + 22
diff --git a/tests/library/exp-string/case.json b/tests/library/exp-string/case.json
index a558ad6..e1ed21a 100644
--- a/tests/library/exp-string/case.json
+++ b/tests/library/exp-string/case.json
@@ -1 +1 @@
-{"sources":["program.cps","../../../pseudocode-lib/_exp_string.cps"]}
+{"sources":["program.cps","../../../pseudocode-lib/_exp_string.cps"],"compiler_args":["-no-std"]}
diff --git a/tests/run.pl b/tests/run.pl
index 8565805..86eff4f 100644
--- a/tests/run.pl
+++ b/tests/run.pl
@@ -443,7 +443,7 @@ sub main {
                     write_file($target, $sentinel);
                 }
                 my $result = run_command($commands,
-                    [$cpsc_abs, @sources, @flags, '-o', $target], $work, '', 120);
+                    [$cpsc_abs, @sources, @{$config->{compiler_args} // []}, @flags, '-o', $target], $work, '', 120);
                 my $preservation_error;
                 if ($existing && (!-f $target || read_file($target) ne $sentinel)) {
                     $preservation_error = 'compilation failure overwrote existing output';
@@ -541,12 +541,12 @@ sub main {
                     my $target = File::Spec->catfile($work,
                         $IS_WINDOWS ? 'program.exe' : 'program.bin');
                     my $compiled = run_command($commands,
-                        [$cpsc_abs, @sources, '-o', $target], $work, '', 120);
+                        [$cpsc_abs, @sources, @{$config->{compiler_args} // []}, '-o', $target], $work, '', 120);
                     check_result($compiled, '');
                     $command = [$target];
                 }
                 else {
-                    $command = [$cpsc_abs, @sources, '-r'];
+                    $command = [$cpsc_abs, @sources, @{$config->{compiler_args} // []}, '-r'];
                 }
                 my $exec_timeout = exists($config->{timeout}) ? $config->{timeout} + 0 : 30;
                 my $result = run_command($commands, $command, $work, $data, $exec_timeout);
diff --git a/tools/driver/StandardLibrary.cc b/tools/driver/StandardLibrary.cc
new file mode 100644
index 0000000..ac88745
--- /dev/null
+++ b/tools/driver/StandardLibrary.cc
@@ -0,0 +1,43 @@
+#include "StandardLibrary.h"
+#include "llvm/Support/FileSystem.h"
+#include "llvm/Support/Path.h"
+#include "llvm/Support/raw_ostream.h"
+
+namespace {
+struct LibraryFile {
+    const char *Name;
+    const char *Data;
+    size_t Size;
+};
+#include "StandardLibrary.inc"
+}
+
+StandardLibrary::~StandardLibrary() {
+    if (!Directory.empty()) llvm::sys::fs::remove_directories(Directory);
+}
+
+bool StandardLibrary::load(std::vector<std::string> &Inputs) {
+    auto Error = llvm::sys::fs::createUniqueDirectory("cpsc-std", Directory);
+    if (Error) {
+        llvm::errs() << "cpsc: fatal error: cannot extract standard library: " << Error.message() << "\n";
+        return false;
+    }
+    for (const auto &File : LibraryFiles) {
+        llvm::SmallString<256> Path(Directory);
+        llvm::sys::path::append(Path, File.Name);
+        llvm::raw_fd_ostream Output(Path, Error);
+        if (!Error) {
+            Output.write(File.Data, File.Size);
+            Output.close();
+            Error = Output.error();
+        }
+        if (Error) {
+            llvm::errs() << "cpsc: fatal error: cannot extract standard library file '"
+                         << File.Name << "': " << Error.message() << "\n";
+            Output.clear_error();
+            return false;
+        }
+        if (llvm::sys::path::extension(Path) == ".cps") Inputs.emplace_back(Path);
+    }
+    return true;
+}
diff --git a/tools/driver/StandardLibrary.h b/tools/driver/StandardLibrary.h
new file mode 100644
index 0000000..5aa4f8c
--- /dev/null
+++ b/tools/driver/StandardLibrary.h
@@ -0,0 +1,12 @@
+#pragma once
+#include "llvm/ADT/SmallString.h"
+#include <string>
+#include <vector>
+
+// Extracted files live through native linking and runner execution.
+class StandardLibrary {
+    llvm::SmallString<256> Directory;
+public:
+    ~StandardLibrary();
+    bool load(std::vector<std::string> &Inputs);
+};
diff --git a/tools/driver/embed-standard-library.pl b/tools/driver/embed-standard-library.pl
new file mode 100644
index 0000000..68be7ef
--- /dev/null
+++ b/tools/driver/embed-standard-library.pl
@@ -0,0 +1,28 @@
+#!/usr/bin/env perl
+use strict;
+use warnings;
+use File::Basename qw(basename);
+
+my ($output, @inputs) = @ARGV;
+open my $out, '>', $output or die "$output: $!\n";
+print {$out} "// Generated from pseudocode-lib; do not edit.\n";
+for my $index (0 .. $#inputs) {
+    open my $in, '<:raw', $inputs[$index] or die "$inputs[$index]: $!\n";
+    local $/;
+    my $data = <$in>;
+    close $in or die "$inputs[$index]: $!\n";
+    print {$out} "static const char LibraryData$index\[\] =\n";
+    while (length $data) {
+        my $chunk = substr($data, 0, 32, '');
+        print {$out} '"', join('', map { sprintf '\\%03o', $_ } unpack 'C*', $chunk), "\"\n";
+    }
+    print {$out} ";\n";
+}
+print {$out} "static const LibraryFile LibraryFiles[] = {\n";
+for my $index (0 .. $#inputs) {
+    my $name = basename($inputs[$index]);
+    die "Invalid library filename: $name\n" unless $name =~ /^[A-Za-z0-9_.-]+$/;
+    print {$out} "    {\"$name\", LibraryData$index, sizeof(LibraryData$index) - 1},\n";
+}
+print {$out} "};\n";
+close $out or die "$output: $!\n";
diff --git a/tools/driver/main.cc b/tools/driver/main.cc
index 5f3aaa2..4d41e5e 100644
--- a/tools/driver/main.cc
+++ b/tools/driver/main.cc
@@ -3,6 +3,7 @@
 #include "cps/Parser.h"
 #include "cps/CodeGen.h"
 #include "NativeCompiler.h"
+#include "StandardLibrary.h"
 #ifdef CPSC_RUN_EMBEDDED
 #include "EmbeddedRunner.h"
 #endif
@@ -91,12 +92,15 @@ int main(int argc, char **argv) {
     std::string OutputPath;
     bool Run = false;
     bool EmitIR = false;
+    bool NoStd = false;
     for (int I = 1; I < argc; ++I) {
         std::string Arg = argv[I];
         if (Arg == "-r") {
             Run = true;
         } else if (Arg == "-S") {
             EmitIR = true;
+        } else if (Arg == "-no-std") {
+            NoStd = true;
         } else if (Arg == "-o") {
             if (!OutputPath.empty()) {
                 llvm::errs() << "cpsc: error: -o may only be specified once\n";
@@ -137,6 +141,9 @@ int main(int argc, char **argv) {
 #endif
     }
     const bool Save = !Run || EmitIR || !OutputPath.empty();
+    const size_t UserInputCount = InputPaths.size();
+    StandardLibrary StdLibrary;
+    if (!NoStd && !StdLibrary.load(InputPaths)) return compilationTerminated();
     std::vector<std::vector<std::unique_ptr<cps::StmtAST>>> Programs(InputPaths.size());
     std::vector<cps::SourceUnit> Units;
     for (size_t I = 0; I < InputPaths.size(); ++I) {
@@ -163,6 +170,9 @@ int main(int argc, char **argv) {
         if (Lex.hadError() || Parser.hadError()) return compilationTerminated();
         Units.push_back({InputPath, Programs[I], std::vector<bool>(InputPaths.size(), false), Lex.isLibrary()});
         Units.back().Visible[I] = true;
+        if (!NoStd && I < UserInputCount)
+            for (size_t J = UserInputCount; J < InputPaths.size(); ++J)
+                Units.back().Visible[J] = true;
     }
     for (size_t I = 0; I < Units.size(); ++I) {
         cps::DiagnosticSource = Units.size() > 1 ? Units[I].Path : "";