Fix typos in comments and documentation (#1072)

Correct "adressing", "constains", "informations", "Aleady" and
"interpretor".

.github/workflows/release.yml

@@ -34,16 +34,16 @@ jobs:
         include:
           - platform: TEGRAX1
             type: Release
-            os: ubuntu-20.04
+            os: ubuntu-22.04
           - platform: TEGRAX1
             type: Trace
-            os: ubuntu-20.04
+            os: ubuntu-22.04
           - platform: TEGRAX1ARM32
             type: Release
-            os: ubuntu-20.04
+            os: ubuntu-22.04
           - platform: TEGRAX1ARM32
             type: Trace
-            os: ubuntu-20.04
+            os: ubuntu-22.04
 
     runs-on: ${{ matrix.os }}
     steps:
@@ -111,7 +111,7 @@ jobs:
           ctest -j$(nproc)
 
       - name: "Upload Artifact"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
           name: box86-${{ matrix.platform }}-${{ matrix.type }}
           path: build/box86

CMakeLists.txt

@@ -20,9 +20,9 @@ option(PHYTIUM "Set to ON if targeting an Phytium (D2000 or FT2000/4) based devi
 option(GAMESHELL "Set to ON if targeting a GameShell device" ${GAMESHELL})
 option(ODROIDXU4 "Set to ON if targeting an ODROID-XU4 device" ${ODROIDXU4})
 option(POWER9 "Set to ON if targeting a POWER9 processor" ${POWER9})
-option(SD845 "Set to ON if targeting a Snapragon 845 based device" ${SD845})
-option(SD865 "Set to ON if targeting a Snapragon 865 based device" ${SD865})
-option(SD888 "Set to ON if targeting a Snapragon 888 based device" ${SD888})
+option(SD845 "Set to ON if targeting a Snapdragon 845 based device" ${SD845})
+option(SD865 "Set to ON if targeting a Snapdragon 865 based device" ${SD865})
+option(SD888 "Set to ON if targeting a Snapdragon 888 based device" ${SD888})
 option(ADLINK "Set to ON if targeting an ADLink AmpereAltra based device" ${ADLINK})
 option(A64 "Set to ON if targeting an Allwinner A64 based device" ${A64})
 option(ARM64 "Set to ON if targeting a generic ARM64 based device" ${ARM64})
@@ -413,6 +413,7 @@ set(WRAPPEDS
     "${BOX86_ROOT}/src/wrapped/wrappedlibbsd.c"
     "${BOX86_ROOT}/src/wrapped/wrappedlibgl.c"
     "${BOX86_ROOT}/src/wrapped/wrappedlibglu.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibglx.c"
     "${BOX86_ROOT}/src/wrapped/wrappedlibx11.c"
     "${BOX86_ROOT}/src/wrapped/wrappedlibasound.c"
     "${BOX86_ROOT}/src/wrapped/wrappedlibdl.c"

debian/changelog

@@ -1,3 +1,18 @@
+box86 (0.3.8) unstable; urgency=low
+* Some optimisation of x86 P flags handling
+* Some x87 fixes and improvments
+* Added some of the tim64 wrapped functions to libc and friends
+* Changed the installation folder of x86 libs to avoid conflict with linux distro
+* Some fixes to internal memory tracking
+* Some more Vulkan extensions wrapped
+* Better handling of BOX86_MAXCPU
+* Some changes and improvments on Signal handling
+* More wrapped functions, and fix to some existing ones
+* A few more syscall, and fix to some existing ones
+* New build profile for SD865
+
+-- Sebastien Chevalier <[email protected]>  Fri, 06 Dec 2024 14:06:00 -0100
+
 box86 (0.3.6) unstable; urgency=low
 * Wrapping: More libs and and function wrapping (gtk3, vulkan...)
 * Reworked X11 Callback handling, for better stability

debian/rules

@@ -10,4 +10,9 @@ override_dh_auto_configure:
 		  -DNOGIT=1
 
 override_dh_shlibdeps:
-	dh_shlibdeps --exclude=libgcc_s.so.1 --exclude=libpng12.so.0 --exclude=libstdc++.so.5 --exclude=libstdc++.so.6 
+	dh_shlibdeps \
+		--exclude=libgcc_s.so.1 \
+		--exclude=libpng12.so.0 \
+		--exclude=libstdc++.so.5 \
+		--exclude=libstdc++.so.6 \
+		--exclude=libunwind.so.8

docs/USAGE.md

@@ -293,6 +293,11 @@ Use of the new fuext_waitc syscall
  * XXX=yyyy
  will add XXX=yyyy env. var. and continue with BOX86_ENV2 ... until var doesn't exist
 
+#### BOX86_ARG0
+Define argv[0] of the executable.
+ * yyyy
+Will rewrite argv[0] instead of using program path.
+
 #### BOX86_FIX_64BIT_INODES *
 Fix/Don't fix 64bit inodes
  * 0 : Don't fix 64bit inodes. (Default.)
@@ -316,7 +321,7 @@ Will use yyyy as x86 bash to launch script. yyyy needs to be a full path to a va
 #### BOX86_BOX64
 Define where to find box64 binary
  * yyyy
-Will use yyyy as x86_64 interpretor, to launch x64_64 binaries
+Will use yyyy as x86_64 interpreter, to launch x64_64 binaries
 
 #### BOX86_JITGDB *
 

postinst

@@ -1,4 +1,6 @@
 #!/bin/sh
+# Restart the systemd-binfmt service to load new binary format rules.
+# This ensures the system is immediately ready to run programs with Box86.
 if [ -d '/run/systemd/system' ]; then
     systemctl restart systemd-binfmt || :
 fi

rebuild_printer.py

@@ -729,7 +729,7 @@ def main(root, ver, __debug_forceAllDebugging=False):
 				
 				variables["u"] = -1 # Unsigned shift
 				
-				variables["w"] = -1 # wback (used in adressing)
+				variables["w"] = -1 # wback (used in addressing)
 				
 				variables["shift"] = False # Shift already detected?
 				

rebuild_wrappers.py

@@ -74,7 +74,7 @@ For each of them:
   NOTE: Those structure letters are "fake types" that are accepted in the macros.
 
 `gbl` contains the first list, `redirects` the second and
- `filespec` constains file specific informations (eg, structures, typedefs required...).
+ `filespec` contains file specific information (eg, structures, typedefs required...).
 
 After sorting the data, it generates:
 
@@ -631,7 +631,7 @@ class Clauses:
 						combinable[len(last_combined) - 1] = True
 						add = (i | attempt[0], m | (i ^ attempt[0]))
 						if add in combined:
-							continue # Aleady added
+							continue # Already added
 						combined.append(add)
 				last_combined.pop()
 			last_combined = combined
@@ -753,6 +753,7 @@ def readFiles(files: Iterable[str]) -> Tuple[JumbledFunctions, JumbledFunctions,
 							 or match("appindicator","appindicator3")\
 							 or match("appindicator","ayatanaappindicator3")\
 							 or match("appindicator3","ayatanaappindicator3")\
+							 or match("libgl",       "libglx")      \
 							 \
 							 or match("libc",        "tcmallocminimal") \
 							 or match("libc",        "ldlinux") 	\
@@ -1150,7 +1151,7 @@ def generate_files(root: str, files: Iterable[str], ver: str, gbls: SortedGlobal
 	for fhdr in files_header:
 		files_guard[fhdr] = trim(files_guard[fhdr])
 	
-	# TODO: put the informations below in a structure (like in box64, where
+	# TODO: put the information below in a structure (like in box64, where
 	# typedefs & co. are stored in an ABI-specific structure)
 	
 	return_x87: str = "DKdf"

src/box86version.h

@@ -3,6 +3,6 @@
 
 #define BOX86_MAJOR 0
 #define BOX86_MINOR 3
-#define BOX86_REVISION 8
+#define BOX86_REVISION 9
 
 #endif //__BOX86_VERSION_H_

src/custommem.c

@@ -999,6 +999,9 @@ void allocProtection(uintptr_t addr, size_t size, uint32_t prot)
     // don't need to add precise tracking probably
 }
 
+uintptr_t pbrk = 0;
+uintptr_t old_brk = 0;
+uintptr_t* cur_brk = NULL;
 void loadProtectionFromMap()
 {
     if(box86_mapclean)
@@ -1015,8 +1018,14 @@ void loadProtectionFromMap()
         if(sscanf(buf, "%x-%x %c%c%c", &s, &e, &r, &w, &x)==5) {
             int prot = ((r=='r')?PROT_READ:0)|((w=='w')?PROT_WRITE:0)|((x=='x')?PROT_EXEC:0);
             allocProtection(s, e-s, prot);
+            if(!pbrk && strstr(buf, "[heap]"))
+                pbrk = s;
         }
     }
+    if(!pbrk) {
+        printf_log(LOG_INFO, "BOX86: Warning, program break not found\n");
+        if(cur_brk) pbrk = *cur_brk;    // approximate is better than nothing
+    }
     fclose(f);
     box86_mapclean = 1;
 }
@@ -1051,12 +1060,18 @@ int getMmapped(uintptr_t addr)
 #define MEDIAN (void*)0x40000000
 static void* findBlockHinted(void* hint, size_t size, uintptr_t mask)
 {
+    // first, check if program break as changed
+    if(pbrk && cur_brk && *cur_brk!=old_brk) {
+        old_brk = *cur_brk;
+        setProtection(pbrk, old_brk-pbrk, PROT_READ|PROT_WRITE);
+    }
     int prot;
     if(hint<LOWEST) hint = LOWEST;
     uintptr_t bend = 0;
     uintptr_t cur = (uintptr_t)hint;
     if(!mask) mask = 0xffff;
-    while(bend!=0xffffffffLL) {
+    uintptr_t end_mem = 0xffffffffLL - mask;
+    while(bend<end_mem) {
         if(!rb_get_end(mapallmem, cur, &prot, &bend)) {
             if(bend-cur>=size)
                 return (void*)cur;
@@ -1165,6 +1180,7 @@ void init_custommem_helper(box86context_t* ctx)
     if(inited) // already initialized
         return;
     inited = 1;
+    cur_brk = dlsym(RTLD_NEXT, "__curbrk");
     memprot = init_rbtree();
     init_mutexes();
 #ifdef DYNAREC

src/dynarec/dynarec_arm_00.c

@@ -2520,13 +2520,25 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                         INST_NAME("REP MOVSB");
                         TSTS_REG_LSL_IMM5(xECX, xECX, 0);
                         B_NEXT(cEQ);    // end of loop
-                        GETDIR(x3,1);
                         SMREAD();
+                        GETDIR(x3, 1);
+                        SUBS_REG_LSL_IMM5(x2, xEDI, xESI, 0);
+                        SUBS_IMM8(x2, x2, 4);
+                        B_MARK(cCC);
+                        MARK2;
+                        SUBS_IMM8(x2, xECX, 4);
+                        B_MARK(cCC);
+                        LDRAI_REG_LSL_IMM5(x1, xESI, x3, 2);
+                        STRAI_REG_LSL_IMM5(x1, xEDI, x3, 2);
+                        MOV_REG(xECX, x2);
+                        B_MARK2(cNE);
+                        B_MARK3(c__);
                         MARK;
                         LDRBAI_REG_LSL_IMM5(x1, xESI, x3, 0);
                         STRBAI_REG_LSL_IMM5(x1, xEDI, x3, 0);
                         SUBS_IMM8(xECX, xECX, 1);
                         B_MARK(cNE);
+                        MARK3;
                         SMWRITE();
                         // done
                         break;

src/dynarec/dynarec_arm_0f.c

@@ -475,7 +475,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                 VMOVD(d1, v0);
             }
             VCMP_F32(d1*2, s0);
-            FCOMI(x1, x2, 0, 0, d1*2, s0, 1);
+            FCOMI(x1, x2);
             break;
 
         case 0x31:

src/dynarec/dynarec_arm_660f.c

@@ -282,7 +282,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
             v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
             GETEX(q0, 0);
             VCMP_F64(v0, q0);
-            FCOMI(x1, x2, 0, 0, v0, q0, 0);
+            FCOMI(x1, x2);
             break;
 
         case 0x38:  // SSSE3 opcodes

src/dynarec/dynarec_arm_d8.c

@@ -100,7 +100,7 @@ uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             break;
         case 0xD8:
         case 0xD9:
@@ -118,7 +118,7 @@ uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             break;
         case 0xE0:
@@ -275,7 +275,7 @@ uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                         VCVT_F64_F32(d1, s0);
                         VCMP_F64(v1, d1);
                     }
-                    FCOM(x1, x2, x3, x14, v1, ST_IS_F(0)?s0:d1, ST_IS_F(0));
+                    FCOM(x1, x2);
                     break;
                 case 3:
                     INST_NAME("FCOMP ST0, float[ED]");
@@ -296,7 +296,7 @@ uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                         VCVT_F64_F32(d1, s0);
                         VCMP_F64(v1, d1);
                     }
-                    FCOM(x1, x2, x3, x14, v1, ST_IS_F(0)?s0:d1, ST_IS_F(0));
+                    FCOM(x1, x2);
                     X87_POP_OR_FAIL(dyn, ninst, x3);
                     break;
                 case 4:

src/dynarec/dynarec_arm_d9.c

@@ -126,7 +126,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64_0(v1);
             }
-            FCOM(x1, x2, 0, 0, v1, 0, ST_IS_F(0));   // same flags...
+            FCOM(x1, x2);   // same flags...
             break;
         case 0xE5:
             INST_NAME("FXAM");

src/dynarec/dynarec_arm_da.c

@@ -128,7 +128,7 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             break;
@@ -185,7 +185,7 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                     VMOVtoV(s0, ed);
                     VCVT_F64_S32(d0, s0);
                     VCMP_F64(v1, d0);
-                    FCOM(x1, x2, x3, x14, v1, d0, 0);
+                    FCOM(x1, x2);
                     break;
                 case 3:
                     INST_NAME("FICOMP ST0, Ed");
@@ -196,7 +196,7 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                     VMOVtoV(s0, ed);
                     VCVT_F64_S32(d0, s0);
                     VCMP_F64(v1, d0);
-                    FCOM(x1, x2, x3, x14, v1, d0, 0);
+                    FCOM(x1, x2);
                     X87_POP_OR_FAIL(dyn, ninst, x3);
                     break;
                 case 4:

src/dynarec/dynarec_arm_db.c

@@ -149,7 +149,7 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOMI(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOMI(x1, x2);
             break;
         case 0xF0:  
         case 0xF1:
@@ -168,7 +168,7 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOMI(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOMI(x1, x2);
             break;
 
         case 0xE0:

src/dynarec/dynarec_arm_dc.c

@@ -96,7 +96,7 @@ uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             break;
         case 0xD8:
         case 0xD9:
@@ -114,7 +114,7 @@ uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             break;
         case 0xE0:
@@ -258,7 +258,7 @@ uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                         VMOVtoV_D(d1, x2, x3);
                     }
                     VCMP_F64(v1, d1);
-                    FCOM(x1, x2, x3, x14, v1, d1, 0);
+                    FCOM(x1, x2);
                     break;
                 case 3:
                     INST_NAME("FCOMP ST0, double[ED]");
@@ -275,7 +275,7 @@ uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                         VMOVtoV_D(d1, x2, x3);
                     }
                     VCMP_F64(v1, d1);
-                    FCOM(x1, x2, x3, x14, v1, d1, 0);
+                    FCOM(x1, x2);
                     X87_POP_OR_FAIL(dyn, ninst, x3);
                     break;
                 case 4:

src/dynarec/dynarec_arm_dd.c

@@ -32,6 +32,7 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
     uint8_t ed;
     int v1, v2;
     int i1, i2, i3;
+    int j32;
 
     MAYUSE(v2);
     MAYUSE(v1);
@@ -111,7 +112,7 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             break;
         case 0xE8:
         case 0xE9:
@@ -129,7 +130,7 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             break;
 
@@ -178,10 +179,53 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                     break;
                 case 1:
                     INST_NAME("FISTTP i64, ST0");
-                    x87_forget(dyn, ninst, x1, x2, 0);
+                    parity = getedparity(dyn, ninst, addr, nextop, 3);
                     addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
-                    if(ed!=x1) {MOV_REG(x1, ed);}
-                    CALL(arm_fistt64, -1, 0);
+                    if(ninst
+                      && dyn->insts[ninst-1].x86.addr
+                      && *(uint8_t*)dyn->insts[ninst-1].x86.addr==0xDF
+                      && (((*(uint8_t*)(dyn->insts[ninst-1].x86.addr+1))>>3)&7)==5)
+                    {
+                        if(parity) {
+                            STRD_IMM8(x2, ed, 0);    // x2/x3 is 64bits
+                        } else {
+                            STR_IMM9(x2, ed, 0);
+                            STR_IMM9(x3, ed, 4);
+                        }
+                    } else {
+                        v1 = x87_get_st(dyn, ninst, x2, x3, 0, NEON_CACHE_ST_D);
+                        //addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
+                        fpu_get_scratch_double(dyn); // to alocate v0
+                        v2 = fpu_get_scratch_double(dyn);
+                        //  get TOP
+                        LDR_IMM9(x14, xEmu, offsetof(x86emu_t, top));
+                        int a = 0 - dyn->n.x87stack;
+                        if(a<0) {
+                            SUB_IMM8(x14, x14, -a);
+                            AND_IMM8(x14, x14, 7);    // (emu->top + i)&7
+                        } else if(a>0) {
+                            ADD_IMM8(x14, x14, a);
+                            AND_IMM8(x14, x14, 7);    // (emu->top + i)&7
+                        }
+                        ADD_REG_LSL_IMM5(x14, xEmu, x14, 4);  // each fpu_ll is 2 int64: ref than ll
+                        MOVW(x2, offsetof(x86emu_t, fpu_ll));   //can be optimized?
+                        ADD_REG_LSL_IMM5(x14, x14, x2, offsetof(fpu_ll_t, sref));
+                        VLDR_64(v2, x14, 0);
+                        VCEQ_32(v2, v2, v1);    // compare
+                        VMOVfrV_D(x2, x3, v2);
+                        ANDS_REG_LSL_IMM5(x2, x2, x3, 0);   // if NE then values are the same!
+                        B_MARK(cEQ);    // do the i64 conversion
+                        // memcpy(ed, &STll(0).ll, sizeof(int64_t));
+                        LDRD_IMM8(x2, x14, offsetof(fpu_ll_t, sq));  // load ll
+                        B_MARK3(c__);
+                        MARK;
+                        MOV32(x2, arm_fist64_3);
+                        VMOV_64(0, v1);    // prepare call to log2
+                        CALL_1DR_U64(x2, x2, x3, x14, (1<<x1));
+                        MARK3;
+                        STR_IMM9(x2, ed, 0);
+                        STR_IMM9(x3, ed, 4);
+                    }
                     X87_POP_OR_FAIL(dyn, ninst, x3);
                     break;
                 case 2:

src/dynarec/dynarec_arm_de.c

@@ -94,7 +94,7 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             break;
 
@@ -107,7 +107,7 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOM(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOM(x1, x2);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             break;

src/dynarec/dynarec_arm_df.c

@@ -90,7 +90,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOMI(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOMI(x1, x2);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             break;
         case 0xF0:
@@ -111,7 +111,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
             } else {
                 VCMP_F64(v1, v2);
             }
-            FCOMI(x1, x2, x3, x14, v1, v2, ST_IS_F(0));
+            FCOMI(x1, x2);
             X87_POP_OR_FAIL(dyn, ninst, x3);
             break;
 
@@ -186,9 +186,9 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                     BIC_IMM8(x3, x3, 0b10011111, 0);
                     VMSR(x3);
                     if(ST_IS_F(0)) {
-                        VCVTR_S32_F32(s0, v1);
+                        VCVT_S32_F32(s0, v1);
                     } else {
-                        VCVTR_S32_F64(s0, v1);
+                        VCVT_S32_F64(s0, v1);
                     }
                     VMRS(x3);   // get the FPCSR reg and test FPU execption (invalid operation only)
                     VMOVfrV(ed, s0);
@@ -337,12 +337,10 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                             STR_IMM9(x3, ed, 4);
                         }
                     } else {
-                        #if 1
                         v1 = x87_get_st(dyn, ninst, x2, x3, 0, NEON_CACHE_ST_D);
                         //addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
                         fpu_get_scratch_double(dyn); // to alocate v0
                         v2 = fpu_get_scratch_double(dyn);
-                        #if 1
                         //  get TOP
                         LDR_IMM9(x14, xEmu, offsetof(x86emu_t, top));
                         int a = 0 - dyn->n.x87stack;
@@ -375,74 +373,6 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                         MARK3;
                         STR_IMM9(x2, ed, 0);
                         STR_IMM9(x3, ed, 4);
-                        #else
-                        v0 = fpu_get_scratch_double(dyn);
-                        s0 = fpu_get_scratch_single(dyn);
-                        // check STll(0).ref==ST(0).q so emu->fpu_ll[emu->top].ref == emu->mmx87[emu->top]
-                        //  get TOP
-                        LDR_IMM9(x14, xEmu, offsetof(x86emu_t, top));
-                        int a = 0 - dyn->n.x87stack;
-                        if(a<0) {
-                            SUB_IMM8(x14, x14, -a);
-                            AND_IMM8(x14, x14, 7);    // (emu->top + i)&7
-                        } else if(a>0) {
-                            ADD_IMM8(x14, x14, a);
-                            AND_IMM8(x14, x14, 7);    // (emu->top + i)&7
-                        }
-                        ADD_REG_LSL_IMM5(x14, xEmu, x14, 4);  // each fpu_ll is 2 int64: ref than ll
-                        MOVW(x2, offsetof(x86emu_t, fpu_ll));   //can be optimized?
-                        ADD_REG_LSL_IMM5(x14, x14, x2, 0);
-                        VLDR_64(v2, x14, 0);
-                        VCEQ_32(v2, v2, v1);    // compare
-                        VMOVfrV_D(x2, x3, v2);
-                        ANDS_REG_LSL_IMM5(x2, x2, x3, 0);   // if NE then values are the same!
-                        B_MARK(cEQ);    // do the i64 conversion
-                        // memcpy(ed, &STll(0).ll, sizeof(int64_t));
-                        LDRD_IMM8(x2, x14, 8);  // load ll
-                        B_MARK3(c__);
-                        MARK;
-                        VEOR(v0, v0, v0);
-                        MOVW(x2, 0x41F0);
-                        VMOVtoDx_16(v0, 3, x2); // V0 = (1<<32) as double
-                        VMOVfrDx_32(x14, v1, 1);    // get high part to extract sign
-                        VABS_F64(v1 ,v1);    //ST0 will be poped, so lost...
-                        VDIV_F64(v2, v1, v0);   // v2 = abs(ST0)/(1<<32) : so 32 bits high part
-                        MSR_nzcvq_0();
-                        VMRS(x2);               // get fpscr
-                        ORR_IMM8(x3, x2, 0b010, 9); // enable exceptions
-                        BIC_IMM8(x3, x3, 0b10011111, 0);
-                        VMSR(x3);
-                        VCVT_U32_F64(s0, v2);   // convert high part to U32
-                        VMRS(x3);   // get the FPCSR reg and test FPU execption (invalid operation only)
-                        VMSR(x2);               // put back fpscr
-                        TSTS_IMM8_ROR(x3, 0b00000001, 0);
-                        B_MARK2(cEQ);   // not overflow...
-                        MARKLOCK;
-                        MOV_IMM(x3, 0b10, 1);   // 0x80000000
-                        MOVW(x2, 0);
-                        B_MARK3(c__);
-                        MARK2;  // continue conversion, it fits an int64!
-                        VCVT_F64_U32(v2, s0);   // int part now
-                        VMLS_F64(v1, v2, v0);   // compute low part
-                        VMOVfrV(x3, s0);        // transfert high path
-                        TSTS_IMM8_ROR(x14, 0b10, 1);    // test high part with 0x800000000
-                        B_MARKLOCK(cNE);        // int overflow...
-                        VCVT_U32_F64(s0, v1);   // convert low part
-                        VMOVfrV(x2, s0);        // transfert low part
-                        TSTS_IMM8_ROR(x14, 0b10, 1);    // 0x800000000
-                        B_MARK3(cEQ);
-                        RSBS_IMM8(x2, x2, 0);   // NEG(i64)
-                        RSC_IMM8(x3, x3, 0);
-                        MARK3;
-                        STR_IMM9(x2, x1, 0);
-                        STR_IMM9(x3, x1, 4);
-                        #endif
-                        #else
-                        MESSAGE(LOG_DUMP, "Need Optimization\n");
-                        x87_forget(dyn, ninst, x2, x3, 0);
-                        if(ed!=x1) {MOV_REG(x1, ed);}
-                        CALL(arm_fistp64, -1, 0);
-                        #endif
                     }
                     X87_POP_OR_FAIL(dyn, ninst, x3);
                     break;

src/dynarec/dynarec_arm_functions.c

@@ -126,21 +126,6 @@ void arm_fbstp(x86emu_t* emu, uint8_t* ed)
     fpu_fbst(emu, ed);
 }
 
-void arm_fistp64(x86emu_t* emu, int64_t* ed)
-{
-    // used of memcpy to avoid aligments issues
-    if(STll(0).sref==ST(0).sq) {
-        memcpy(ed, &STll(0).sq, sizeof(int64_t));
-    } else {
-        int64_t tmp;
-        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(ST0.d))
-            tmp = 0x8000000000000000LL;
-        else
-            tmp = fpu_round(emu, ST0.d);
-        memcpy(ed, &tmp, sizeof(tmp));
-    }
-}
-
 int64_t arm_fist64_0(double d)
 {
     int64_t tmp;
@@ -178,13 +163,6 @@ int64_t arm_fist64_3(double d)
     return tmp;
 }
 
-void arm_fistt64(x86emu_t* emu, int64_t* ed)
-{
-    // used of memcpy to avoid aligments issues
-    int64_t tmp = ST0.d;
-    memcpy(ed, &tmp, sizeof(tmp));
-}
-
 void arm_fld(x86emu_t* emu, uint8_t* ed)
 {
     memcpy(&STld(0).ld, ed, 10);

src/dynarec/dynarec_arm_functions.h

@@ -16,8 +16,6 @@ void arm_fscale(x86emu_t* emu);
 void arm_fbld(x86emu_t* emu, uint8_t* ed);
 void arm_fild64(x86emu_t* emu, int64_t* ed);
 void arm_fbst(x86emu_t* emu, uint8_t* ed);
-void arm_fistp64(x86emu_t* emu, int64_t* ed);
-void arm_fistt64(x86emu_t* emu, int64_t* ed);
 void arm_fld(x86emu_t* emu, uint8_t* ed);
 void arm_fsave(x86emu_t* emu, uint8_t* ed);
 void arm_frstor(x86emu_t* emu, uint8_t* ed);

src/dynarec/dynarec_arm_helper.h

@@ -316,7 +316,7 @@
 #define IFXN(A, B)  if((dyn->insts[ninst].x86.gen_flags&(A) && !(dyn->insts[ninst].x86.gen_flags&(B))))
 
 // Generate FCOM with s1 and s2 scratch regs (the VCMP is already done)
-#define FCOM(s1, s2, s3, s4, v1, v2, is_f)                                      \
+#define FCOM(s1, s2)                                                            \
     VMRS_APSR();    /* 0b0100011100000000 */                                    \
     LDRH_IMM8(s1, xEmu, offsetof(x86emu_t, sw));   /*offset is 8bits right?*/   \
     BIC_IMM8(s1, s1, 0b01000111, 12);                                           \
@@ -324,36 +324,10 @@
     ORR_IMM8_COND(cEQ, s1, s1, 0b01000000, 12); /* equal */                     \
     ORR_IMM8_COND(cMI, s1, s1, 0b00000001, 12); /* less than */                 \
     /* greater than leave 0 */                                                  \
-    if(s4) {                                                                    \
-        Bcond(cVS, (is_f?11:13)*4-8);                                           \
-        if(is_f) {                                                              \
-            MOVW(s4, 0);                                                        \
-            MOVT(s4, 0x7ff0); /* +inf */                                        \
-            VMOVfrV(s2, v1);                                                    \
-            CMPS_REG_LSL_IMM5(s2, s4, 0);                                       \
-            Bcond(cEQ, 5*4-8); /* same */                                       \
-            VMOVfrV(s2, v2);                                                    \
-            ORR_IMM8(s4, s4, 0b10, 1); /* -inf */                               \
-            CMPS_REG_LSL_IMM5(s2, s4, 0);                                       \
-        } else {                                                                \
-            MOVW(s4, 0);                                                        \
-            MOVT(s4, 0x7ff0); /* +inf */                                        \
-            VMOVfrV_D(s2, s3, v1);                                              \
-            ORR_REG_LSL_IMM5(s2, s2, s3, 0);                                    \
-            CMPS_REG_LSL_IMM5(s2, s4, 0);                                       \
-            Bcond(cEQ, 6*4-8); /* same */                                       \
-            VMOVfrV_D(s2, s3, v2);                                              \
-            ORR_REG_LSL_IMM5(s2, s2, s3, 0);                                    \
-            ORR_IMM8(s4, s4, 0b10, 1); /* -inf */                               \
-            CMPS_REG_LSL_IMM5(s2, s4, 0);                                       \
-        }                                                                       \
-        Bcond(cNE, 4+4-8); /* same */                                           \
-        MOVW(s1, 0);                                                            \
-    }                                                                           \
     STRH_IMM8(s1, xEmu, offsetof(x86emu_t, sw))
 
 // Generate FCOMI with s1 and s2 scratch regs (the VCMP is already done)
-#define FCOMI(s1, s2, s3, s4, v1, v2, is_f)                                 \
+#define FCOMI(s1, s2)                                                       \
     IFX(X_CF|X_PF|X_ZF|X_PEND) {                                            \
         VMRS_APSR();    /* 0b111 */                                         \
         BIC_IMM8(xFlags, xFlags, 0b1000101, 0);                             \
@@ -361,38 +335,10 @@
         ORR_IMM8_COND(cEQ, xFlags, xFlags, 0b01000000, 0); /* zero */       \
         ORR_IMM8_COND(cMI, xFlags, xFlags, 0b00000001, 0); /* less than */  \
         /* greater than leave 0 */                                          \
-        if(s4) {                                                            \
-            Bcond(cVS, (is_f?11:13)*4-8);                                   \
-            if(is_f) {                                                      \
-                MOVW(s4, 0);                                                \
-                MOVT(s4, 0x7ff0); /* +inf */                                \
-                VMOVfrV(s2, v1);                                            \
-                CMPS_REG_LSL_IMM5(s2, s4, 0);                               \
-                Bcond(cEQ, 5*4-8); /* same */                               \
-                VMOVfrV(s2, v2);                                            \
-                ORR_IMM8(s4, s4, 0b10, 1); /* -inf */                       \
-                CMPS_REG_LSL_IMM5(s2, s4, 0);                               \
-            } else {                                                        \
-                MOVW(s4, 0);                                                \
-                MOVT(s4, 0x7ff0); /* +inf */                                \
-                VMOVfrV_D(s2, s3, v1);                                      \
-                ORR_REG_LSL_IMM5(s2, s2, s3, 0);                            \
-                CMPS_REG_LSL_IMM5(s2, s4, 0);                               \
-                Bcond(cEQ, 6*4-8); /* same */                               \
-                VMOVfrV_D(s2, s3, v2);                                      \
-                ORR_REG_LSL_IMM5(s2, s2, s3, 0);                            \
-                ORR_IMM8(s4, s4, 0b10, 1); /* -inf */                       \
-                CMPS_REG_LSL_IMM5(s2, s4, 0);                               \
-            }                                                               \
-            Bcond(cNE, 4+4-8); /* same */                                   \
-            BIC_IMM8(xFlags, xFlags, 0b1000101, 0);                         \
-        }                                                                   \
-    }                                                                       \
-    SET_DFNONE(s1);                                                         \
-    IFX(X_OF|X_AF|X_SF|X_PEND) {                                            \
         /* clear 0b100010010000 */                                          \
         BIC_IMM8(xFlags, xFlags, 0b10001001, 14);                           \
     }                                                                       \
+    SET_DFNONE(s1);                                                         \
 
 
 #if STEP == 0

src/elfs/elfloader.c

@@ -1808,7 +1808,7 @@ static int my_dl_iterate_phdr_##A(struct dl_phdr_info* a, size_t b, void* c)
 {                                                                                   \
     if(!a->dlpi_name)                                                               \
         return 0;                                                                   \
-    if(!a->dlpi_name[0]) /*don't send informations about box86 itself*/             \
+    if(!a->dlpi_name[0]) /*don't send information about box86 itself*/             \
         return 0;                                                                   \
     return RunFunctionFmt(my_dl_iterate_phdr_fct_##A, "pLp", a, b, c);  \
 }
@@ -2033,6 +2033,9 @@ kh_defaultversion_t* GetWeakDefaultVersion(elfheader_t* h)
     return h?h->weakdefver:NULL;
 }
 
+#ifndef STT_GNU_IFUNC
+#define STT_GNU_IFUNC 10
+#endif
 
 uintptr_t pltResolver = ~(uintptr_t)0;
 EXPORT void PltResolver(x86emu_t* emu)
@@ -2076,6 +2079,11 @@ EXPORT void PltResolver(x86emu_t* emu)
             printf_dump(LOG_DEBUG, "symbol %s from %s but elf not initialized yet, run Init now (from %s)\n", symname, ElfName(sym_elf), ElfName(h));
             RunElfInitPltResolver(sym_elf, emu);
         }
+        if(sym_elf && sym && ((sym->st_info&0xf)==STT_GNU_IFUNC)) {
+            // this is an IFUNC, needs to evaluate the function first!
+            printf_dump(LOG_DEBUG, "            Indirect function, will call the resolver now at %p\n", (void*)offs);
+            offs = RunFunction(offs, 0);
+        }
         offs = (uintptr_t)getAlternate((void*)offs);
 
         if(p) {

src/emu/modrm.h

@@ -108,7 +108,7 @@
             case 7: base =      R_BX; break;    \
         }                                       \
         switch((m>>6)&3) {                      \
-            case 0: if(m&7==6) base = F16S; break; \
+            case 0: if((m&7)==6) base = F16S; break; \
             case 1: base += F8S; break;         \
             case 2: base += F16S; break;        \
         }                                       \

src/emu/x86int3.c

@@ -53,12 +53,12 @@ x86emu_t* x86emu_fork(x86emu_t* emu, int forktype)
         // error...
     } else if(v!=0) {  
         // execute atforks parent functions
-        for (int i=0; i<my_context->atfork_sz; --i)
+        for (int i=0; i<my_context->atfork_sz; ++i)
             if(my_context->atforks[i].parent)
                 EmuCall(emu, my_context->atforks[i].parent);
     } else if(v==0) {
         // execute atforks child functions
-        for (int i=0; i<my_context->atfork_sz; --i)
+        for (int i=0; i<my_context->atfork_sz; ++i)
             if(my_context->atforks[i].child)
                 EmuCall(emu, my_context->atforks[i].child);
     }

src/emu/x86run0f.c

@@ -270,7 +270,7 @@ uintptr_t Run0F(x86emu_t *emu, uintptr_t addr, int *step)
                 GM.sd[1] = 0x80000000;
             else
                 GM.sd[1] = EX->f[1];
-            if(isnanf(EX->f[0]) || isinff(EX->f[0]) || EX->f[0]>=(float)0x80000000U || EX->f[1]<-(float)0x80000000U)
+            if(isnanf(EX->f[0]) || isinff(EX->f[0]) || EX->f[0]>=(float)0x80000000U || EX->f[0]<-(float)0x80000000U)
                 GM.sd[0] = 0x80000000;
             else
                 GM.sd[0] = EX->f[0];

src/emu/x86run64.c

@@ -12,6 +12,7 @@
 #include "x86trace.h"
 #include "box86context.h"
 #include "bridge.h"
+#include "signals.h"
 #ifdef DYNAREC
 #include "../dynarec/arm_lock_helper.h"
 #endif
@@ -416,6 +417,40 @@ uintptr_t Run64(x86emu_t *emu, int seg, uintptr_t addr)
             return addr-1;       // ignore FS: to execute regular opcode
             break;
 
+        case 0xF7:                      /* GRP3 Ed(,Id) */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            switch((nextop>>3)&7) {
+                case 0: 
+                case 1:                 /* TEST Ed,Id */
+                    tmp32u = F32;
+                    test32(emu, ED->dword[0], tmp32u);
+                    break;
+                case 2:                 /* NOT Ed */
+                    ED->dword[0] = not32(emu, ED->dword[0]);
+                    break;
+                case 3:                 /* NEG Ed */
+                    ED->dword[0] = neg32(emu, ED->dword[0]);
+                    break;
+                case 4:                 /* MUL EAX,Ed */
+                    mul32_eax(emu, ED->dword[0]);
+                    break;
+                case 5:                 /* IMUL EAX,Ed */
+                    imul32_eax(emu, ED->dword[0]);
+                    break;
+                case 6:                 /* DIV Ed */
+                    if(!ED->dword[0])
+                        emit_div0(emu, (void*)R_EIP, 0);
+                    div32(emu, ED->dword[0]);
+                    break;
+                case 7:                 /* IDIV Ed */
+                    if(!ED->dword[0])
+                        emit_div0(emu, (void*)R_EIP, 0);
+                    idiv32(emu, ED->dword[0]);
+                    break;
+            }
+            break;
+
         case 0xFF:              /* GRP 5 Ed */
             nextop = F8;
             GET_ED_OFFS(tlsdata);

src/emu/x86run6466.c

@@ -48,6 +48,12 @@ uintptr_t Run6466(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr)
             GW.word[0] = EW->word[0];
             break;
 
+        case 0x8D:                      	    /* LEA Gw,M */
+            nextop = F8;
+            GET_ED_;    // segment doesn't enter in the computation
+            GW.word[0] = (uint16_t)(uint32_t)ED;
+            break;
+
         default:
             return 0;
     }

src/emu/x86run66.c

@@ -284,7 +284,11 @@ uintptr_t Run66(x86emu_t *emu, int rep, uintptr_t addr)
             GET_EW;
             EW->word[0] = emu->segs[(nextop&0x38)>>3];
             break;
-        
+        case 0x8D:                      	/* LEA Gw,M */
+            nextop = F8;
+            GET_ED_;
+            GW.word[0] = (uint16_t)(uint32_t)ED;
+            break;
         case 0x8E:                               /* MOV Seg,Ew */
             nextop = F8;
             GET_EW;
@@ -584,4 +588,4 @@ uintptr_t Run66(x86emu_t *emu, int rep, uintptr_t addr)
             return 0;
     }
     return addr;
-}
+}

src/emu/x86runf066.c

@@ -147,7 +147,7 @@ uintptr_t RunF066(x86emu_t *emu, uintptr_t addr)
                         } while(arm_lock_write_h(ED, inc16(emu, tmp16u)));
 #else
                     pthread_mutex_lock(&emu->context->mutex_lock);
-                    ED->dword[0] = inc32(emu, ED->dword[0]);
+                    ED->word[0] = inc16(emu, ED->word[0]);
                     pthread_mutex_unlock(&emu->context->mutex_lock);
 #endif
                     break;
@@ -168,7 +168,7 @@ uintptr_t RunF066(x86emu_t *emu, uintptr_t addr)
                         } while(arm_lock_write_h(ED, dec16(emu, tmp16u)));
 #else
                     pthread_mutex_lock(&emu->context->mutex_lock);
-                    ED->dword[0] = dec32(emu, ED->dword[0]);
+                    ED->word[0] = dec16(emu, ED->word[0]);
                     pthread_mutex_unlock(&emu->context->mutex_lock);
 #endif
                     break;

src/emu/x87emu_private.h

@@ -84,10 +84,6 @@ static inline void fpu_fcom(x86emu_t* emu, double b)
         emu->sw.f.F87_C0 = 1;
         emu->sw.f.F87_C2 = 1;
         emu->sw.f.F87_C3 = 1;
-    } else if ((b==-INFINITY) || (ST0.d==INFINITY)) {
-        emu->sw.f.F87_C0 = 0;
-        emu->sw.f.F87_C2 = 0;
-        emu->sw.f.F87_C3 = 0;
     } else if (isgreater(ST0.d, b)) {
         emu->sw.f.F87_C0 = 0;
         emu->sw.f.F87_C2 = 0;
@@ -115,10 +111,6 @@ static inline void fpu_fcomi(x86emu_t* emu, double b)
         SET_FLAG(F_CF);
         SET_FLAG(F_PF);
         SET_FLAG(F_ZF);
-    } else if ((b==-INFINITY) || (ST0.d==INFINITY)) {
-        CLEAR_FLAG(F_CF);
-        CLEAR_FLAG(F_PF);
-        CLEAR_FLAG(F_ZF);
     } else if (isgreater(ST0.d, b)) {
         CLEAR_FLAG(F_CF);
         CLEAR_FLAG(F_PF);

src/librarian/library.c

@@ -380,14 +380,14 @@ static void initDummyLib(library_t *lib)
 
 static const char* essential_libs[] = {
     "libc.so.6", "libpthread.so.0", "librt.so.1", "libGL.so.1", "libGL.so", "libX11.so.6", 
-    "libasound.so.2", "libdl.so.2", "libm.so.6",
+    "libasound.so.2", "libdl.so.2", "libm.so.6", "libbsd.so.0", "libbsd.so", "ld-linux.so.2",
     "libXxf86vm.so.1", "libXinerama.so.1", "libXrandr.so.2", "libXext.so.6", "libXfixes.so.3", "libXcursor.so.1",
     "libXrender.so.1", "libXft.so.2", "libXi.so.6", "libXss.so.1", "libXpm.so.4", "libXau.so.6", "libXdmcp.so.6",
     "libX11-xcb.so.1", "libxcb.so.1", "libxcb-xfixes.so.0", "libxcb-shape.so.0", "libxcb-shm.so.0", "libxcb-randr.so.0",
     "libxcb-image.so.0", "libxcb-keysyms.so.1", "libxcb-xtest.so.0", "libxcb-glx.so.0", "libxcb-dri2.so.0", "libxcb-dri3.so.0",
     "libXtst.so.6", "libXt.so.6", "libXcomposite.so.1", "libXdamage.so.1", "libXmu.so.6", "libxkbcommon.so.0", 
     "libxkbcommon-x11.so.0", "libpulse-simple.so.0", "libpulse.so.0", "libvulkan.so.1", "libvulkan.so",
-    "ld-linux-x86-64.so.2", "crashhandler.so", "libtcmalloc_minimal.so.0", "libtcmalloc_minimal.so.4"
+    "crashhandler.so", "libtcmalloc_minimal.so.0", "libtcmalloc_minimal.so.4"
 };
 static int isEssentialLib(const char* name) {
     for (int i=0; i<sizeof(essential_libs)/sizeof(essential_libs[0]); ++i)

src/library_list.h

@@ -5,9 +5,12 @@
 GO("libpthread.so.0", libpthread)
 GO("librt.so.1", librt)
 GO("libbsd.so.0", libbsd)
+GO("libbsd.so", libbsd)
 GO("libGL.so.1", libgl)
 GO("libGL.so", libgl)
 GO("libGLU.so.1", libglu)
+GO("libGLX.so.0", libglx)
+GO("libGLX.so", libglx)
 GO("libX11.so.6", libx11)
 GO("libasound.so.2", libasound)
 GO("libSDL2-2.0.so.0", sdl2)
@@ -183,7 +186,6 @@ GO("libpcre.so.3", libpcre)
 #ifdef ANDROID
 GO("libc.so", libc)
 GO("libm.so", libm)
-GO("libbsd.so", libbsd)
 GO("libdl.so", libdl)
 GO("libcups.so", libcups)
 GO("liblber-2.4.so", lber)

src/libtools/stat64_helper.c

@@ -1,6 +1,8 @@
 #define _TIME_BITS 64
 #define _FILE_OFFSET_BITS 64
-
+#ifdef PANDORA
+#define _ATFILE_SOURCE
+#endif
 #include <string.h>
 #include <sys/stat.h>
 

src/main.c

@@ -1459,8 +1459,10 @@ int main(int argc, const char **argv, char **env)
         my_context->box64path = getenv("BOX86_BOX64");
         if(!my_context->box64path) {
             my_context->box64path = box_strdup(my_context->box86path);
-            char* p = strrchr(my_context->box64path, '8');  // get the 8 of box86
-            p[0] = '6'; p[1] = '4'; // change 86 to 64
+            if (strstr(my_context->box64path, "box86")) {
+                char* p = strrchr(my_context->box64path, '8');  // get the 8 of box86
+                p[0] = '6'; p[1] = '4'; // change 86 to 64
+            }
         }
         if(!FileExist(my_context->box64path, IS_FILE)) {
             box_free(my_context->box64path);

src/mallochook.c

@@ -692,7 +692,7 @@ EXPORT void* my_safer_scalable_aligned_realloc(void* p, size_t size, size_t alig
     void* new_p = box_memalign(align, size);
     memcpy(new_p, p, (old_size<size)?old_size:size);
     box_free(p);
-    return p;
+    return new_p;
 }
 
 EXPORT void my_safer_scalable_free(void*p , void* old)
@@ -740,7 +740,7 @@ EXPORT void* my_scalable_aligned_realloc(void* p, size_t size, size_t align)
     void* new_p = box_memalign(align, size);
     memcpy(new_p, p, (old_size<size)?old_size:size);
     box_free(p);
-    return p;
+    return new_p;
 }
 
 EXPORT size_t my_scalable_msize(void* p)

src/wrapped/generated/functions_list.txt

@@ -187,6 +187,7 @@
 #() iFuu
 #() iFup
 #() iFli
+#() iFLL
 #() iFLp
 #() iFpw
 #() iFpi
@@ -3381,6 +3382,10 @@ wrappedlibglu:
   - gluNurbsCallback
   - gluQuadricCallback
   - gluTessCallback
+wrappedlibglx:
+- pFp:
+  - glXGetProcAddress
+  - glXGetProcAddressARB
 wrappedlibharfbuzz:
 wrappedlibibus:
 - vFpippp:

src/wrapped/generated/wrappedlibglxdefs.h

@@ -0,0 +1,8 @@
+/*******************************************************************
+ * File automatically generated by rebuild_wrappers.py (v2.0.0.11) *
+ *******************************************************************/
+#ifndef __wrappedlibglxDEFS_H_
+#define __wrappedlibglxDEFS_H_
+
+
+#endif // __wrappedlibglxDEFS_H_

src/wrapped/generated/wrappedlibglxtypes.h

@@ -0,0 +1,20 @@
+/*******************************************************************
+ * File automatically generated by rebuild_wrappers.py (v2.0.0.11) *
+ *******************************************************************/
+#ifndef __wrappedlibglxTYPES_H_
+#define __wrappedlibglxTYPES_H_
+
+#ifndef LIBNAME
+#error You should only #include this file inside a wrapped*.c file
+#endif
+#ifndef ADDED_FUNCTIONS
+#define ADDED_FUNCTIONS() 
+#endif
+
+typedef void* (*pFp_t)(void*);
+
+#define SUPER() ADDED_FUNCTIONS() \
+	GO(glXGetProcAddress, pFp_t) \
+	GO(glXGetProcAddressARB, pFp_t)
+
+#endif // __wrappedlibglxTYPES_H_

src/wrapped/generated/wrappedlibglxundefs.h

@@ -0,0 +1,8 @@
+/*******************************************************************
+ * File automatically generated by rebuild_wrappers.py (v2.0.0.11) *
+ *******************************************************************/
+#ifndef __wrappedlibglxUNDEFS_H_
+#define __wrappedlibglxUNDEFS_H_
+
+
+#endif // __wrappedlibglxUNDEFS_H_

src/wrapped/generated/wrapper.c

@@ -259,6 +259,7 @@ typedef int32_t (*iFui_t)(uint32_t, int32_t);
 typedef int32_t (*iFuu_t)(uint32_t, uint32_t);
 typedef int32_t (*iFup_t)(uint32_t, void*);
 typedef int32_t (*iFli_t)(intptr_t, int32_t);
+typedef int32_t (*iFLL_t)(uintptr_t, uintptr_t);
 typedef int32_t (*iFLp_t)(uintptr_t, void*);
 typedef int32_t (*iFpw_t)(void*, int16_t);
 typedef int32_t (*iFpi_t)(void*, int32_t);
@@ -2715,6 +2716,7 @@ void iFui(x86emu_t *emu, uintptr_t fcn) { iFui_t fn = (iFui_t)fcn; R_EAX=fn(*(ui
 void iFuu(x86emu_t *emu, uintptr_t fcn) { iFuu_t fn = (iFuu_t)fcn; R_EAX=fn(*(uint32_t*)(R_ESP + 4), *(uint32_t*)(R_ESP + 8)); }
 void iFup(x86emu_t *emu, uintptr_t fcn) { iFup_t fn = (iFup_t)fcn; R_EAX=fn(*(uint32_t*)(R_ESP + 4), *(void**)(R_ESP + 8)); }
 void iFli(x86emu_t *emu, uintptr_t fcn) { iFli_t fn = (iFli_t)fcn; R_EAX=fn(*(intptr_t*)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); }
+void iFLL(x86emu_t *emu, uintptr_t fcn) { iFLL_t fn = (iFLL_t)fcn; R_EAX=fn(*(uintptr_t*)(R_ESP + 4), *(uintptr_t*)(R_ESP + 8)); }
 void iFLp(x86emu_t *emu, uintptr_t fcn) { iFLp_t fn = (iFLp_t)fcn; R_EAX=fn(*(uintptr_t*)(R_ESP + 4), *(void**)(R_ESP + 8)); }
 void iFpw(x86emu_t *emu, uintptr_t fcn) { iFpw_t fn = (iFpw_t)fcn; R_EAX=fn(*(void**)(R_ESP + 4), *(int16_t*)(R_ESP + 8)); }
 void iFpi(x86emu_t *emu, uintptr_t fcn) { iFpi_t fn = (iFpi_t)fcn; R_EAX=fn(*(void**)(R_ESP + 4), *(int32_t*)(R_ESP + 8)); }

src/wrapped/generated/wrapper.h

@@ -219,6 +219,7 @@ void iFui(x86emu_t *emu, uintptr_t fnc);
 void iFuu(x86emu_t *emu, uintptr_t fnc);
 void iFup(x86emu_t *emu, uintptr_t fnc);
 void iFli(x86emu_t *emu, uintptr_t fnc);
+void iFLL(x86emu_t *emu, uintptr_t fnc);
 void iFLp(x86emu_t *emu, uintptr_t fnc);
 void iFpw(x86emu_t *emu, uintptr_t fnc);
 void iFpi(x86emu_t *emu, uintptr_t fnc);

src/wrapped/wrappedgobject2_private.h

@@ -266,7 +266,7 @@ GO(g_thread_get_type, iFv)
 GO(g_time_zone_get_type, iFv)
 //GOM(g_type_add_class_cache_func, vFEpB)
 GO(g_type_add_class_private, vFpiu)
-//GO(g_type_add_instance_private, 
+GO(g_type_add_instance_private, iFLL)
 //GOM(g_type_add_interface_check, vFEpB)
 //GOM(g_type_add_interface_dynamic, vFEiip)
 GOM(g_type_add_interface_static, vFEiip)

src/wrapped/wrappedlibbsd.c

@@ -19,11 +19,9 @@
 #include "callback.h"
 
 
-#ifdef ANDROID
-const char* libbsdName = "libbsd.so";
-#else
 const char* libbsdName = "libbsd.so.0";
-#endif
+#define ALTNAME "libbsd.so"
+
 #define LIBNAME libbsd
 
 #define PRE_INIT\

src/wrapped/wrappedlibc.c

@@ -2572,7 +2572,6 @@ EXPORT int32_t my_execvp(x86emu_t* emu, const char* path, char* const argv[])
     // fullpath is gone, so the search will only be on PATH, not on BOX86_PATH (is that an issue?)
     return execvp(path, argv);
 }
-
 // execvpe should use PATH to search for the program first
 EXPORT int32_t my_execvpe(x86emu_t* emu, const char* path, char* const argv[], char* const envv[])
 {
@@ -2599,22 +2598,35 @@ EXPORT int32_t my_execvpe(x86emu_t* emu, const char* path, char* const argv[], c
         if(self) newargv[1] = emu->context->fullpath;
         if(script) newargv[2] = emu->context->bashpath;
         printf_log(LOG_DEBUG, " => execvp(\"%s\", %p [\"%s\", \"%s\"...:%d])\n", newargv[0], newargv, newargv[1], i?newargv[2]:"", i);
+        #ifdef PANDORA
+        int ret = execve(newargv[0], newargv, envv);
+        #else
         int ret = execvpe(newargv[0], newargv, envv);
+        #endif
         box_free(fullpath);
         return ret;
     }
+    #ifndef PANDORA
     box_free(fullpath);
+    #endif
     if((!strcmp(path + strlen(path) - strlen("/uname"), "/uname") || !strcmp(path, "uname"))
      && argv[1] && (!strcmp(argv[1], "-m") || !strcmp(argv[1], "-p") || !strcmp(argv[1], "-i"))
      && !argv[2]) {
         // uname -m is redirected to box86 -m
         path = my_context->box86path;
         char *argv2[3] = { my_context->box86path, argv[1], NULL };
+        #ifdef PANDORA
+        return execve(fullpath, argv2, envv);
+        #else
         return execvpe(path, argv2, envv);
+        #endif
     }
-
+    #ifdef PANDORA
+    return execve(fullpath, argv, envv);
+    #else
     // fullpath is gone, so the search will only be on PATH, not on BOX86_PATH (is that an issue?)
     return execvpe(path, argv, envv);
+    #endif
 }
 
 // execvp should use PATH to search for the program first
@@ -3683,10 +3695,11 @@ EXPORT char my___libc_single_threaded = 0;
     "libm.so",              \
     "libbsd.so"
 #else
-#define NEEDED_LIBS   4,\
+#define NEEDED_LIBS   5,\
     "ld-linux.so.2",    \
     "libpthread.so.0",  \
     "librt.so.1",       \
+    "libm.so.6",        \
     "libbsd.so.0"
 #define NEEDED_LIBS_234 6,  \
     "ld-linux.so.2",        \

src/wrapped/wrappedlibglx.c

@@ -0,0 +1,53 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#define _GNU_SOURCE         /* See feature_test_macros(7) */
+#include <dlfcn.h>
+
+#include "wrappedlibs.h"
+
+#include "debug.h"
+#include "wrapper.h"
+#include "bridge.h"
+#include "librarian/library_private.h"
+#include "x86emu.h"
+#include "emu/x86emu_private.h"
+#include "callback.h"
+#include "librarian.h"
+#include "box86context.h"
+#include "myalign.h"
+#include "gltools.h"
+
+const char* libglxName = "libGLX.so.0";
+#define LIBNAME libglx
+
+#include "generated/wrappedlibglxtypes.h"
+
+#include "wrappercallback.h"
+
+#define SUPER() \
+GO(0)   \
+GO(1)   \
+GO(2)   \
+GO(3)   \
+
+
+#undef SUPER
+
+EXPORT void* myx_glXGetProcAddress(x86emu_t* emu, void* name) 
+{
+    khint_t k;
+    const char* rname = (const char*)name;
+    return getGLProcAddress(emu, (glprocaddress_t)my->glXGetProcAddress, rname);
+}
+
+EXPORT void* myx_glXGetProcAddressARB(x86emu_t* emu, void* name) 
+{
+    khint_t k;
+    const char* rname = (const char*)name;
+    return getGLProcAddress(emu, (glprocaddress_t)my->glXGetProcAddressARB, rname);
+}
+
+#define ALTMY myx_
+
+#include "wrappedlib_init.h"

src/wrapped/wrappedlibglx_private.h

@@ -0,0 +1,45 @@
+#if !(defined(GO) && defined(GOM) && defined(GO2) && defined(DATA))
+#error meh!
+#endif
+
+// __glXGLLoadGLXFunction
+GO(glXChooseFBConfig, pFpipp)
+GO(glXChooseVisual, pFpip)
+GO(glXCopyContext,vFppp)
+GO(glXCreateContext,pFpppi)
+GO(glXCreateGLXPixmap,pFppp)
+GO(glXCreateNewContext,pFppipi)
+GO(glXCreatePbuffer,pFppp)
+GO(glXCreatePixmap,pFppp)
+GO(glXCreateWindow,pFpppp)
+GO(glXDestroyContext,vFpp)
+GO(glXDestroyGLXPixmap,vFpp)
+GO(glXDestroyPbuffer,vFpp)
+GO(glXDestroyPixmap,vFpp)
+GO(glXDestroyWindow,vFpp)
+GO(glXGetClientString, pFpi)
+GO(glXGetConfig, iFppip)
+GO(glXGetCurrentContext, pFv)
+GO(glXGetCurrentDisplay, pFv)
+GO(glXGetCurrentDrawable, pFv)
+GO(glXGetCurrentReadDrawable, pFv)
+GO(glXGetFBConfigAttrib, iFppip)
+GO(glXGetFBConfigs,pFpip)
+GOM(glXGetProcAddress, pFEp)
+GOM(glXGetProcAddressARB, pFEp)
+GO(glXGetSelectedEvent, vFppp)
+GO(glXGetVisualFromFBConfig, pFpp)
+GO(glXIsDirect,iFpp)
+GO(glXMakeContextCurrent,iFpppp)
+GO(glXMakeCurrent,iFppp)
+GO(glXQueryContext,iFppip)
+GO(glXQueryDrawable, iFppip)
+GO(glXQueryExtension, iFppp)
+GO(glXQueryExtensionsString,pFpi)
+GO(glXQueryServerString,pFpii)
+GO(glXQueryVersion,iFppp)
+GO(glXSelectEvent, vFppu)
+GO(glXSwapBuffers,vFpp)
+GO(glXUseXFont,vFpiii)
+GO(glXWaitGL,vFv)
+GO(glXWaitX,vFv)

src/wrapped/wrappedxml2_private.h

@@ -203,7 +203,7 @@ GO(xmlCheckFilename, iFp)
 GO(xmlCheckHTTPInput, pFpp)
 GO(xmlCheckLanguageID, iFp)
 GO(xmlCheckUTF8, iFp)
-//GO(xmlCheckVersion, 
+GO(xmlCheckVersion,  vFi)
 GO(xmlChildElementCount, LFp)
 GO(xmlCleanupCharEncodingHandlers, vFv)
 GO(xmlCleanupEncodingAliases, vFv)

tests/test19.c

@@ -32,7 +32,7 @@ uint64_t _fucompp_(double a, double b)
     :"=a" (ret):"m"(a), "m"(b):"cc");
     return ret;
 }
-uint64_t _fistpw_(double a)
+uint64_t _fistpw_(double a, uint8_t rd)
 {
     uint16_t ret;
     uint16_t t1, t2;
@@ -40,15 +40,31 @@ uint64_t _fistpw_(double a)
     "fldl %1\n"
     "fstcw %2\n"
     "mov %2, %%ax\n"
-    "mov $0x0c, %%ah\n"
+    "mov %4, %%ah\n"
     "mov %%ax, %3\n"
     "fldcw %3\n"
     "fistp %0\n"
     "fldcw %2\n"
-    :"=m" (ret):"m"(a), "m"(t1), "m"(t2):"cc");
+    :"=m" (ret):"m"(a), "m"(t1), "m"(t2), "m"(rd):"cc");
     return ret;
 }
-uint64_t _fistpl_(double a)
+uint64_t _fisttpw_(double a, uint8_t rd)
+{
+    uint16_t ret;
+    uint16_t t1, t2;
+    asm volatile (
+    "fldl %1\n"
+    "fstcw %2\n"
+    "mov %2, %%ax\n"
+    "mov %4, %%ah\n"
+    "mov %%ax, %3\n"
+    "fldcw %3\n"
+    "fisttp %0\n"
+    "fldcw %2\n"
+    :"=m" (ret):"m"(a), "m"(t1), "m"(t2), "m"(rd):"cc");
+    return ret;
+}
+uint64_t _fistpl_(double a, uint8_t rd)
 {
     uint32_t ret;
     uint16_t t1, t2;
@@ -56,15 +72,31 @@ uint64_t _fistpl_(double a)
     "fldl %1\n"
     "fstcw %2\n"
     "mov %2, %%ax\n"
-    "mov $0x0c, %%ah\n"
+    "mov %4, %%ah\n"
     "mov %%ax, %3\n"
     "fldcw %3\n"
     "fistpl %0\n"
     "fldcw %2\n"
-    :"=m" (ret):"m"(a), "m"(t1), "m"(t2):"cc");
+    :"=m" (ret):"m"(a), "m"(t1), "m"(t2), "m"(rd):"cc");
     return ret;
 }
-uint64_t _fistpq_(double a)
+uint64_t _fisttpl_(double a, uint8_t rd)
+{
+    uint32_t ret;
+    uint16_t t1, t2;
+    asm volatile (
+    "fldl %1\n"
+    "fstcw %2\n"
+    "mov %2, %%ax\n"
+    "mov %4, %%ah\n"
+    "mov %%ax, %3\n"
+    "fldcw %3\n"
+    "fisttpl %0\n"
+    "fldcw %2\n"
+    :"=m" (ret):"m"(a), "m"(t1), "m"(t2), "m"(rd):"cc");
+    return ret;
+}
+uint64_t _fistpq_(double a, uint8_t rd)
 {
     uint64_t ret;
     uint16_t t1, t2;
@@ -72,15 +104,15 @@ uint64_t _fistpq_(double a)
     "fldl %1\n"
     "fstcw %2\n"
     "mov %2, %%ax\n"
-    "mov $0x0c, %%ah\n"
+    "mov %4, %%ah\n"
     "mov %%ax, %3\n"
     "fldcw %3\n"
     "fistpq %0\n"
     "fldcw %2\n"
-    :"=m" (ret):"m"(a), "m"(t1), "m"(t2):"cc");
+    :"=m" (ret):"m"(a), "m"(t1), "m"(t2), "m"(rd):"cc");
     return ret;
 }
-uint64_t _frndint_(double a)
+uint64_t _fisttpq_(double a, uint8_t rd)
 {
     uint64_t ret;
     uint16_t t1, t2;
@@ -88,13 +120,29 @@ uint64_t _frndint_(double a)
     "fldl %1\n"
     "fstcw %2\n"
     "mov %2, %%ax\n"
-    "mov $0x0c, %%ah\n"
+    "mov %4, %%ah\n"
+    "mov %%ax, %3\n"
+    "fldcw %3\n"
+    "fisttpq %0\n"
+    "fldcw %2\n"
+    :"=m" (ret):"m"(a), "m"(t1), "m"(t2), "m"(rd):"cc");
+    return ret;
+}
+uint64_t _frndint_(double a, uint8_t rd)
+{
+    uint64_t ret;
+    uint16_t t1, t2;
+    asm volatile (
+    "fldl %1\n"
+    "fstcw %2\n"
+    "mov %2, %%ax\n"
+    "mov %4, %%ah\n"
     "mov %%ax, %3\n"
     "fldcw %3\n"
     "frndint\n"
     "fstpl %0\n"
     "fldcw %2\n"
-    :"=m" (ret):"m"(a), "m"(t1), "m"(t2):"cc");
+    :"=m" (ret):"m"(a), "m"(t1), "m"(t2), "m"(rd):"cc");
     return ret;
 }
 #endif
@@ -105,44 +153,54 @@ int main(int argc, const char** argv)
  char pf, cf, zf;
  char pa, ca, za;
  uint64_t flags;
- uint64_t tests[][2] = {
-  {0x0, 0x4082c00000000000LL},
-  {0x8000000000000000LL, 0x4082c00000000000LL},
-  {0x8000000000000000LL, 0x0},
-  {0x8000000000000000LL, 0x3ff0000000000000LL},
-  {0x3ff0000000000000LL, 0x3fe89d9000000000LL},
-  {0x3ff0000000000000LL, 0x7ff0000000000000LL},
-  {0xfff0000000000000LL, 0x7ff0000000000000LL},
-  {0x3ff0002ca0000000LL, 0xaeff000025000000LL},
-  {0x3ff0000050000000LL, 0xc082c00000000000LL},
-  {0x0, 0x7ff8000000000001LL},
-  {0x7ff8000000000001LL, 0x7ff8000000000001LL},
- };
- int n = sizeof(tests)/sizeof(tests[0]); 
- for(int i=0; i<n; ++i)
-  for(int j=0; j<2; ++j) {
-   *(uint64_t*)&a = tests[i][0+j];
-   *(uint64_t*)&b = tests[i][1-j];
-   printf("FUCOMI 0x%llx 0x%llx ", *(uint64_t*)&a, *(uint64_t*)&b);
-   flags = _fucomip_(a, b);
-   ca = (flags>>0)&1?'C':'-';
-   za = (flags>>(0+6))&1?'Z':'-';
-   pa = (flags>>(0+2))&1?'P':'-';
-   printf("%c%c%c\n", za, pa, ca);
-   printf("FUCOMPP 0x%llx 0x%llx ", *(uint64_t*)&a, *(uint64_t*)&b);
-   flags = _fucompp_(a, b);
-   ca = (flags>>8)&1?'C':'-';
-   za = (flags>>(8+6))&1?'Z':'-';
-   pa = (flags>>(8+2))&1?'P':'-';
-   printf("%c%c%c\n", za, pa, ca);
-   printf("FRNDINT 0x%llx => 0x%llx\n", *(uint64_t*)&a, _frndint_(a));
-   printf("FRNDINT 0x%llx => 0x%llx\n", *(uint64_t*)&b, _frndint_(b));
-   printf("FISTP 0x%llx => word: %x\n", *(uint64_t*)&a, _fistpw_(a));
-   printf("FISTP 0x%llx => word: %x\n", *(uint64_t*)&b, _fistpw_(b));
-   printf("FISTP 0x%llx => long: %x\n", *(uint64_t*)&a, _fistpl_(a));
-   printf("FISTP 0x%llx => long: %x\n", *(uint64_t*)&b, _fistpl_(b));
-   printf("FISTP 0x%llx => quad: %llx\n", *(uint64_t*)&a, _fistpq_(a));
-   printf("FISTP 0x%llx => quad: %llx\n", *(uint64_t*)&b, _fistpq_(b));
+ uint64_t tests[] = {
+    0x0,
+    0x4082c00000000000LL,
+    0x8000000000000000LL,
+    0x40dfffc000000000LL, 
+    0x40e0002000000000LL, 	// 0x7fff and 0x8001 as double
+    0xc0e0002000000000LL, 
+    0xc0dfffc000000000LL,	// -0x8001 and -0x7fff as double
+    0x3ff0000000000000LL,
+    0x3fe89d9000000000LL,
+    0x7ff0000000000000LL,
+    0xfff0000000000000LL, 
+    0x3ff0002ca0000000LL, 
+    0xaeff000025000000LL,
+    0x3ff0000050000000LL, 
+    0xc082c00000000000LL,
+    0x7ff8000000000001LL,
+};
+int n = sizeof(tests)/sizeof(tests[0]); 
+for(int i=0; i<n; ++i)
+    for(int j=0; j<n; ++j) {
+        *(uint64_t*)&a = tests[i];
+        *(uint64_t*)&b = tests[j];
+        printf("FUCOMI 0x%llx 0x%llx ", *(uint64_t*)&a, *(uint64_t*)&b);
+        flags = _fucomip_(a, b);
+        ca = (flags>>0)&1?'C':'-';
+        za = (flags>>(0+6))&1?'Z':'-';
+        pa = (flags>>(0+2))&1?'P':'-';
+        printf("%c%c%c\n", za, pa, ca);
+        printf("FUCOMPP 0x%llx 0x%llx ", *(uint64_t*)&a, *(uint64_t*)&b);
+        flags = _fucompp_(a, b);
+        ca = (flags>>8)&1?'C':'-';
+        za = (flags>>(8+6))&1?'Z':'-';
+        pa = (flags>>(8+2))&1?'P':'-';
+        printf("%c%c%c\n", za, pa, ca);
+    }
+for(int i=0; i<n; ++i) {
+    *(uint64_t*)&a = tests[i];
+    for(int rd=0; rd<3; ++rd) {
+        printf("Rounding %d\n", rd);
+        printf(" FRNDINT 0x%llx => 0x%llx\n", *(uint64_t*)&a, _frndint_(a, rd<<2));
+        printf(" FISTP 0x%llx => word: %x\n", *(uint64_t*)&a, _fistpw_(a, rd<<2));
+        printf(" FISTP 0x%llx => long: %x\n", *(uint64_t*)&a, _fistpl_(a, rd<<2));
+        printf(" FISTP 0x%llx => quad: %llx\n", *(uint64_t*)&a, _fistpq_(a, rd<<2));
+        printf(" FISTTP 0x%llx => word: %x\n", *(uint64_t*)&a, _fisttpw_(a, rd<<2));
+        printf(" FISTTP 0x%llx => long: %x\n", *(uint64_t*)&a, _fisttpl_(a, rd<<2));
+        printf(" FISTTP 0x%llx => quad: %llx\n", *(uint64_t*)&a, _fisttpq_(a, rd<<2));
+    }
  }
  printf("\nDone\n");
 }