make buildable on Linux

11 files changed, 77 insertions, 87 deletions

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 8286b30..c8ccabc 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -31,21 +31,19 @@ jobs:
   build:
     strategy:
       matrix:
-        toolset: [mingw, vs2022, vs2026]
+        toolset: [gcc, vs2022, vs2026]
         platform: [x64, x86]
         use_asm: [1, 0]
         configuration: [Debug, Release]
         include:
           - boost-version: 1.81.0
-          # MinGW builds are done on Linux, since it's more up-to-date there,
-          # and it's much faster.
-          - {toolset: mingw, os: ubuntu-latest}
+          - {toolset: gcc, os: ubuntu-latest}
           - {toolset: vs2022, os: windows-2022}
           # Boost only supports VS 2026 starting with version 1.91, I think?
           - {toolset: vs2026, os: windows-2025, boost-version: 1.91.0}
         exclude:
           - {platform: x64, use_asm: 1}
-          - {toolset: mingw, use_asm: 1}
+          - {toolset: gcc, use_asm: 1}
     runs-on: '${{ matrix.os }}'
     name: |
       Build: ${{ matrix.toolset }} / ${{ matrix.platform }} / ${{ matrix.use_asm && '+asm' || '-asm' }} / ${{ matrix.configuration }}
@@ -69,11 +67,11 @@ jobs:
         uses: actions/setup-python@v6
         with:
           python-version: '3.x'
-      - name: Set up MinGW
-        uses: egor-tensin/setup-mingw@v3
+      - name: Set up GCC
+        uses: egor-tensin/setup-gcc@v2
         with:
           platform: '${{ matrix.platform }}'
-        if: matrix.toolset == 'mingw'
+        if: matrix.toolset == 'gcc'
       - name: Build Boost
         run: make deps
       - name: Build
@@ -86,4 +84,3 @@ jobs:
           if-no-files-found: error
       - name: Test
         run: make test
-        if: runner.os == 'Windows'
diff --git a/Makefile b/Makefile
index 6081509..46aa80e 100644
--- a/Makefile
+++ b/Makefile
@@ -7,7 +7,7 @@ boost_dir   := $(build_dir)/boost
 cmake_dir   := $(build_dir)/cmake
 install_dir := $(build_dir)/install
 
-TOOLSET         ?= mingw
+TOOLSET         ?= auto
 PLATFORM        ?= auto
 CONFIGURATION   ?= Debug
 BOOST_VERSION   ?= 1.72.0
diff --git a/aes/include/aes/aes.h b/aes/include/aes/aes.h
index 8f1e511..3cfaf8a 100644
--- a/aes/include/aes/aes.h
+++ b/aes/include/aes/aes.h
@@ -9,6 +9,7 @@
 
 #include "block.h"
 #include "round_keys.h"
+#include "workarounds.h"
 
 #include <assert.h>
 
@@ -16,13 +17,13 @@
 extern "C" {
 #endif
 
-AES_Block __fastcall aes128_encrypt_block_internal(AES_Block, const AES128_RoundKeys*);
-AES_Block __fastcall aes192_encrypt_block_internal(AES_Block, const AES192_RoundKeys*);
-AES_Block __fastcall aes256_encrypt_block_internal(AES_Block, const AES256_RoundKeys*);
+AES_Block AES_ASM_ATTR aes128_encrypt_block_internal(AES_Block, const AES128_RoundKeys*);
+AES_Block AES_ASM_ATTR aes192_encrypt_block_internal(AES_Block, const AES192_RoundKeys*);
+AES_Block AES_ASM_ATTR aes256_encrypt_block_internal(AES_Block, const AES256_RoundKeys*);
 
-AES_Block __fastcall aes128_decrypt_block_internal(AES_Block, const AES128_RoundKeys*);
-AES_Block __fastcall aes192_decrypt_block_internal(AES_Block, const AES192_RoundKeys*);
-AES_Block __fastcall aes256_decrypt_block_internal(AES_Block, const AES256_RoundKeys*);
+AES_Block AES_ASM_ATTR aes128_decrypt_block_internal(AES_Block, const AES128_RoundKeys*);
+AES_Block AES_ASM_ATTR aes192_decrypt_block_internal(AES_Block, const AES192_RoundKeys*);
+AES_Block AES_ASM_ATTR aes256_decrypt_block_internal(AES_Block, const AES256_RoundKeys*);
 
 static inline AES_Block aes128_encrypt_block(AES_Block plaintext, const AES128_RoundKeys* keys) {
     assert(keys);
diff --git a/aes/include/aes/round_keys.h b/aes/include/aes/round_keys.h
index be35f24..4d9af69 100644
--- a/aes/include/aes/round_keys.h
+++ b/aes/include/aes/round_keys.h
@@ -10,6 +10,7 @@
 #include "block.h"
 #include "error.h"
 #include "key.h"
+#include "workarounds.h"
 
 #include <assert.h>
 
@@ -29,31 +30,25 @@ typedef struct {
     AES_Block keys[15];
 } AES256_RoundKeys;
 
-void __fastcall aes128_expand_key_internal(AES_Block key, AES128_RoundKeys* encryption_keys);
+void AES_ASM_ATTR aes128_expand_key_internal(AES_Block key, AES128_RoundKeys* encryption_keys);
 
-void __fastcall aes192_expand_key_internal(
-    AES_Block key_lo,
-    AES_Block key_hi,
-    AES192_RoundKeys* encryption_keys
-);
+void AES_ASM_ATTR
+aes192_expand_key_internal(AES_Block key_lo, AES_Block key_hi, AES192_RoundKeys* encryption_keys);
 
-void __fastcall aes256_expand_key_internal(
-    AES_Block key_lo,
-    AES_Block key_hi,
-    AES256_RoundKeys* encryption_keys
-);
+void AES_ASM_ATTR
+aes256_expand_key_internal(AES_Block key_lo, AES_Block key_hi, AES256_RoundKeys* encryption_keys);
 
-void __fastcall aes128_derive_decryption_keys_internal(
+void AES_ASM_ATTR aes128_derive_decryption_keys_internal(
     const AES128_RoundKeys* encryption_keys,
     AES128_RoundKeys* decryption_keys
 );
 
-void __fastcall aes192_derive_decryption_keys_internal(
+void AES_ASM_ATTR aes192_derive_decryption_keys_internal(
     const AES192_RoundKeys* encryption_keys,
     AES192_RoundKeys* decryption_keys
 );
 
-void __fastcall aes256_derive_decryption_keys_internal(
+void AES_ASM_ATTR aes256_derive_decryption_keys_internal(
     const AES256_RoundKeys* encryption_keys,
     AES256_RoundKeys* decryption_keys
 );
diff --git a/aes/include/aes/workarounds.h b/aes/include/aes/workarounds.h
index 4a728c4..084a689 100644
--- a/aes/include/aes/workarounds.h
+++ b/aes/include/aes/workarounds.h
@@ -16,3 +16,12 @@
 #endif
 
 #define AES_UNUSED_PARAMETER(...) (void)(__VA_ARGS__)
+
+/* Targetting 32-bit Windows, match the existing ASM implementation's behavior
+ * (__fastcall). Otherwise, do nothing, using the standard calling convention
+ * by default. */
+#if defined(_MSC_VER) && !defined(_WIN64)
+#define AES_ASM_ATTR __fastcall
+#else
+#define AES_ASM_ATTR
+#endif
diff --git a/aes/src/c/aes128.c b/aes/src/c/aes128.c
index 80892ad..379c4cb 100644
--- a/aes/src/c/aes128.c
+++ b/aes/src/c/aes128.c
@@ -10,10 +10,8 @@
 #include <emmintrin.h>
 #include <wmmintrin.h>
 
-AES_Block __fastcall aes128_encrypt_block_internal(
-    AES_Block plaintext,
-    const AES128_RoundKeys* encryption_keys
-) {
+AES_Block AES_ASM_ATTR
+aes128_encrypt_block_internal(AES_Block plaintext, const AES128_RoundKeys* encryption_keys) {
     plaintext = _mm_xor_si128(plaintext, encryption_keys->keys[0]);
     plaintext = _mm_aesenc_si128(plaintext, encryption_keys->keys[1]);
     plaintext = _mm_aesenc_si128(plaintext, encryption_keys->keys[2]);
@@ -27,10 +25,8 @@ AES_Block __fastcall aes128_encrypt_block_internal(
     return _mm_aesenclast_si128(plaintext, encryption_keys->keys[10]);
 }
 
-AES_Block __fastcall aes128_decrypt_block_internal(
-    AES_Block ciphertext,
-    const AES128_RoundKeys* decryption_keys
-) {
+AES_Block AES_ASM_ATTR
+aes128_decrypt_block_internal(AES_Block ciphertext, const AES128_RoundKeys* decryption_keys) {
     ciphertext = _mm_xor_si128(ciphertext, decryption_keys->keys[0]);
     ciphertext = _mm_aesdec_si128(ciphertext, decryption_keys->keys[1]);
     ciphertext = _mm_aesdec_si128(ciphertext, decryption_keys->keys[2]);
@@ -44,7 +40,7 @@ AES_Block __fastcall aes128_decrypt_block_internal(
     return _mm_aesdeclast_si128(ciphertext, decryption_keys->keys[10]);
 }
 
-static AES_Block __fastcall aes128_expand_key_assist(AES_Block prev, AES_Block hwgen) {
+static AES_Block AES_ASM_ATTR aes128_expand_key_assist(AES_Block prev, AES_Block hwgen) {
     AES_Block tmp = prev;
 
     tmp = _mm_slli_si128(tmp, 4);
@@ -60,7 +56,7 @@ static AES_Block __fastcall aes128_expand_key_assist(AES_Block prev, AES_Block h
     return prev;
 }
 
-void __fastcall aes128_expand_key_internal(AES_Block key, AES128_RoundKeys* encryption_keys) {
+void AES_ASM_ATTR aes128_expand_key_internal(AES_Block key, AES128_RoundKeys* encryption_keys) {
     AES_Block prev = encryption_keys->keys[0] = key;
     prev = encryption_keys->keys[1] =
         aes128_expand_key_assist(prev, _mm_aeskeygenassist_si128(prev, 0x01));
@@ -84,7 +80,7 @@ void __fastcall aes128_expand_key_internal(AES_Block key, AES128_RoundKeys* encr
         aes128_expand_key_assist(prev, _mm_aeskeygenassist_si128(prev, 0x36));
 }
 
-void __fastcall aes128_derive_decryption_keys_internal(
+void AES_ASM_ATTR aes128_derive_decryption_keys_internal(
     const AES128_RoundKeys* encryption_keys,
     AES128_RoundKeys* decryption_keys
 ) {
diff --git a/aes/src/c/aes192.c b/aes/src/c/aes192.c
index d3f83fc..83ee354 100644
--- a/aes/src/c/aes192.c
+++ b/aes/src/c/aes192.c
@@ -10,10 +10,8 @@
 #include <emmintrin.h>
 #include <wmmintrin.h>
 
-AES_Block __fastcall aes192_encrypt_block_internal(
-    AES_Block plaintext,
-    const AES192_RoundKeys* encryption_keys
-) {
+AES_Block AES_ASM_ATTR
+aes192_encrypt_block_internal(AES_Block plaintext, const AES192_RoundKeys* encryption_keys) {
     plaintext = _mm_xor_si128(plaintext, encryption_keys->keys[0]);
     plaintext = _mm_aesenc_si128(plaintext, encryption_keys->keys[1]);
     plaintext = _mm_aesenc_si128(plaintext, encryption_keys->keys[2]);
@@ -29,10 +27,8 @@ AES_Block __fastcall aes192_encrypt_block_internal(
     return _mm_aesenclast_si128(plaintext, encryption_keys->keys[12]);
 }
 
-AES_Block __fastcall aes192_decrypt_block_internal(
-    AES_Block ciphertext,
-    const AES192_RoundKeys* decryption_keys
-) {
+AES_Block AES_ASM_ATTR
+aes192_decrypt_block_internal(AES_Block ciphertext, const AES192_RoundKeys* decryption_keys) {
     ciphertext = _mm_xor_si128(ciphertext, decryption_keys->keys[0]);
     ciphertext = _mm_aesdec_si128(ciphertext, decryption_keys->keys[1]);
     ciphertext = _mm_aesdec_si128(ciphertext, decryption_keys->keys[2]);
@@ -48,11 +44,8 @@ AES_Block __fastcall aes192_decrypt_block_internal(
     return _mm_aesdeclast_si128(ciphertext, decryption_keys->keys[12]);
 }
 
-static void __fastcall aes192_expand_key_assist(
-    AES_Block* prev_lo,
-    AES_Block* prev_hi,
-    AES_Block hwgen
-) {
+static void AES_ASM_ATTR
+aes192_expand_key_assist(AES_Block* prev_lo, AES_Block* prev_hi, AES_Block hwgen) {
     AES_Block tmp = *prev_lo;
 
     tmp = _mm_slli_si128(tmp, 4);
@@ -73,11 +66,8 @@ static void __fastcall aes192_expand_key_assist(
     *prev_hi = _mm_xor_si128(*prev_hi, tmp);
 }
 
-void __fastcall aes192_expand_key_internal(
-    AES_Block key_lo,
-    AES_Block key_hi,
-    AES192_RoundKeys* encryption_keys
-) {
+void AES_ASM_ATTR
+aes192_expand_key_internal(AES_Block key_lo, AES_Block key_hi, AES192_RoundKeys* encryption_keys) {
     encryption_keys->keys[0] = key_lo;
     encryption_keys->keys[1] = key_hi;
 
@@ -125,7 +115,7 @@ void __fastcall aes192_expand_key_internal(
     encryption_keys->keys[12] = key_lo;
 }
 
-void __fastcall aes192_derive_decryption_keys_internal(
+void AES_ASM_ATTR aes192_derive_decryption_keys_internal(
     const AES192_RoundKeys* encryption_keys,
     AES192_RoundKeys* decryption_keys
 ) {
diff --git a/aes/src/c/aes256.c b/aes/src/c/aes256.c
index 8a86f40..5a57761 100644
--- a/aes/src/c/aes256.c
+++ b/aes/src/c/aes256.c
@@ -10,10 +10,8 @@
 #include <emmintrin.h>
 #include <wmmintrin.h>
 
-AES_Block __fastcall aes256_encrypt_block_internal(
-    AES_Block plaintext,
-    const AES256_RoundKeys* encryption_keys
-) {
+AES_Block AES_ASM_ATTR
+aes256_encrypt_block_internal(AES_Block plaintext, const AES256_RoundKeys* encryption_keys) {
     plaintext = _mm_xor_si128(plaintext, encryption_keys->keys[0]);
     plaintext = _mm_aesenc_si128(plaintext, encryption_keys->keys[1]);
     plaintext = _mm_aesenc_si128(plaintext, encryption_keys->keys[2]);
@@ -31,10 +29,8 @@ AES_Block __fastcall aes256_encrypt_block_internal(
     return _mm_aesenclast_si128(plaintext, encryption_keys->keys[14]);
 }
 
-AES_Block __fastcall aes256_decrypt_block_internal(
-    AES_Block ciphertext,
-    const AES256_RoundKeys* decryption_keys
-) {
+AES_Block AES_ASM_ATTR
+aes256_decrypt_block_internal(AES_Block ciphertext, const AES256_RoundKeys* decryption_keys) {
     ciphertext = _mm_xor_si128(ciphertext, decryption_keys->keys[0]);
     ciphertext = _mm_aesdec_si128(ciphertext, decryption_keys->keys[1]);
     ciphertext = _mm_aesdec_si128(ciphertext, decryption_keys->keys[2]);
@@ -52,11 +48,8 @@ AES_Block __fastcall aes256_decrypt_block_internal(
     return _mm_aesdeclast_si128(ciphertext, decryption_keys->keys[14]);
 }
 
-static AES_Block __fastcall aes256_expand_key_assist(
-    AES_Block* prev_lo,
-    AES_Block* prev_hi,
-    AES_Block hwgen
-) {
+static AES_Block AES_ASM_ATTR
+aes256_expand_key_assist(AES_Block* prev_lo, AES_Block* prev_hi, AES_Block hwgen) {
     AES_Block tmp = *prev_lo;
 
     tmp = _mm_slli_si128(tmp, 4);
@@ -75,11 +68,8 @@ static AES_Block __fastcall aes256_expand_key_assist(
     return *prev_hi;
 }
 
-void __fastcall aes256_expand_key_internal(
-    AES_Block key_lo,
-    AES_Block key_hi,
-    AES256_RoundKeys* encryption_keys
-) {
+void AES_ASM_ATTR
+aes256_expand_key_internal(AES_Block key_lo, AES_Block key_hi, AES256_RoundKeys* encryption_keys) {
     AES_Block prev_lo, prev_hi;
     AES_Block hwgen;
 
@@ -139,7 +129,7 @@ void __fastcall aes256_expand_key_internal(
     encryption_keys->keys[14] = aes256_expand_key_assist(&prev_lo, &prev_hi, hwgen);
 }
 
-void __fastcall aes256_derive_decryption_keys_internal(
+void AES_ASM_ATTR aes256_derive_decryption_keys_internal(
     const AES256_RoundKeys* encryption_keys,
     AES256_RoundKeys* decryption_keys
 ) {
diff --git a/aesxx/CMakeLists.txt b/aesxx/CMakeLists.txt
index 704231e..21ed76e 100644
--- a/aesxx/CMakeLists.txt
+++ b/aesxx/CMakeLists.txt
@@ -3,7 +3,10 @@ set(CMAKE_CXX_STANDARD_REQUIRED ON)
 
 add_library(aesxx INTERFACE)
 target_include_directories(aesxx INTERFACE include/)
-target_link_libraries(aesxx INTERFACE aes dbghelp)
+target_link_libraries(aesxx INTERFACE aes)
+if(WIN32)
+    target_link_libraries(aesxx INTERFACE dbghelp)
+endif()
 
 if(MSVC_VERSION EQUAL 1900)
     # These annoying dbghelp.h warnings:
diff --git a/aesxx/utils/CMakeLists.txt b/aesxx/utils/CMakeLists.txt
index 44c4acc..8b904a6 100644
--- a/aesxx/utils/CMakeLists.txt
+++ b/aesxx/utils/CMakeLists.txt
@@ -35,13 +35,17 @@ set(file_util_headers
 add_util(encrypt_file encrypt_file.cpp ${file_util_headers})
 add_util(decrypt_file decrypt_file.cpp ${file_util_headers})
 
-set(bmp_util_headers ${file_util_headers} helpers/bmp.hpp)
+if(WIN32)
+    set(bmp_util_headers ${file_util_headers} helpers/bmp.hpp)
 
-add_util(encrypt_bmp encrypt_bmp.cpp ${bmp_util_headers})
-add_util(decrypt_bmp decrypt_bmp.cpp ${bmp_util_headers})
+    add_util(encrypt_bmp encrypt_bmp.cpp ${bmp_util_headers})
+    add_util(decrypt_bmp decrypt_bmp.cpp ${bmp_util_headers})
+endif()
 
 find_package(Python3 REQUIRED COMPONENTS Interpreter)
 
 include(test/basic.cmake)
 include(test/block.cmake)
-include(test/bmp.cmake)
+if(WIN32)
+    include(test/bmp.cmake)
+endif()
diff --git a/aesxx/utils/test/basic.cmake b/aesxx/utils/test/basic.cmake
index b8df5a3..cba1180 100644
--- a/aesxx/utils/test/basic.cmake
+++ b/aesxx/utils/test/basic.cmake
@@ -1,11 +1,16 @@
+list(APPEND utils block file)
+if(WIN32)
+    list(APPEND utils bmp)
+endif()
+
 foreach(action encrypt decrypt)
-    foreach(util block file bmp)
+    foreach(util ${utils})
         set(exe "${action}_${util}")
         set(tgt "util_${exe}")
         add_test(NAME "${tgt}_help" COMMAND Python3::Interpreter
             "${CMAKE_SOURCE_DIR}/cmake/tools/ctest-driver.py"
             run
-            --pass-regex "usage: ${exe}\\.exe" "set algorithm"
+            --pass-regex "usage: ${exe}" "set algorithm"
             --fail-regex "usage error:"
             --
             "$<TARGET_FILE:${tgt}>" -h
@@ -16,7 +21,7 @@ foreach(action encrypt decrypt)
             --exit-code 1
             --pass-regex
                 "usage error: the option '--algorithm' is required"
-                "usage: ${exe}\\.exe"
+                "usage: ${exe}"
                 "set algorithm"
             --
             "$<TARGET_FILE:${tgt}>"