py/nlr: Fix nlr functions for 64bit ports built with gcc on Windows

The number of registers used should be 10, not 12, to match the assembly
code in nlrx64.c. With this change the 64bit mingw builds don't need to
use the setjmp implementation, and this fixes miscellaneous crashes and
assertion failures as reported in #1751 for instance.

To avoid mistakes in the future where something gcc-related for Windows
only gets fixed for one particular compiler/environment combination,
make use of a MICROPY_NLR_OS_WINDOWS macro.

To make sure everything nlr-related is now ok when built with gcc this
has been verified with:
- unix port built with gcc on Cygwin (i686-pc-cygwin-gcc and
  x86_64-pc-cygwin-gcc, version 6.4.0)
- windows port built with mingw-w64's gcc from Cygwin
 (i686-w64-mingw32-gcc and x86_64-w64-mingw32-gcc, version 6.4.0)
 and MSYS2 (like the ones on Cygwin but version 7.2.0)

ports/windows/Makefile

@@ -50,10 +50,6 @@ CFLAGS_MOD += -DMICROPY_USE_READLINE=2
 LDFLAGS_MOD += -lreadline
 endif
 
-ifeq ($(CROSS_COMPILE),x86_64-w64-mingw32-)
-CFLAGS_MOD += -DMICROPY_NLR_SETJMP=1
-endif
-
 LIB += -lws2_32
 
 # List of sources for qstr extraction

py/nlr.c

@@ -28,7 +28,7 @@
 
 #if !MICROPY_NLR_SETJMP
 // When not using setjmp, nlr_push_tail is called from inline asm so needs special c
-#if MICROPY_NLR_X86 && (defined(_WIN32) || defined(__CYGWIN__))
+#if MICROPY_NLR_X86 && MICROPY_NLR_OS_WINDOWS
 // On these 32-bit platforms make sure nlr_push_tail doesn't have a leading undersco
 unsigned int nlr_push_tail(nlr_buf_t *nlr) asm("nlr_push_tail");
 #else

py/nlr.h

@@ -36,13 +36,19 @@
 
 // If MICROPY_NLR_SETJMP is not enabled then auto-detect the machine arch
 #if !MICROPY_NLR_SETJMP
+// A lot of nlr-related things need different treatment on Windows
+#if defined(_WIN32) || defined(__CYGWIN__)
+#define MICROPY_NLR_OS_WINDOWS 1
+#else
+#define MICROPY_NLR_OS_WINDOWS 0
+#endif
 #if defined(__i386__)
     #define MICROPY_NLR_X86 (1)
     #define MICROPY_NLR_NUM_REGS (6)
 #elif defined(__x86_64__)
     #define MICROPY_NLR_X64 (1)
-    #if defined(__CYGWIN__)
-        #define MICROPY_NLR_NUM_REGS (12)
+    #if MICROPY_NLR_OS_WINDOWS
+        #define MICROPY_NLR_NUM_REGS (10)
     #else
         #define MICROPY_NLR_NUM_REGS (8)
     #endif

py/nlrx64.c

@@ -33,18 +33,12 @@
 // x86-64 callee-save registers are:
 //  rbx, rbp, rsp, r12, r13, r14, r15
 
-#if defined(_WIN32) || defined(__CYGWIN__)
-#define NLR_OS_WINDOWS 1
-#else
-#define NLR_OS_WINDOWS 0
-#endif
-
 __attribute__((used)) unsigned int nlr_push_tail(nlr_buf_t *nlr);
 
 unsigned int nlr_push(nlr_buf_t *nlr) {
     (void)nlr;
 
-    #if NLR_OS_WINDOWS
+    #if MICROPY_NLR_OS_WINDOWS
 
     __asm volatile (
     "movq   (%rsp), %rax        \n" // load return %rip
@@ -93,7 +87,7 @@ NORETURN void nlr_jump(void *val) {
 
     __asm volatile (
     "movq   %0, %%rcx           \n" // %rcx points to nlr_buf
-    #if NLR_OS_WINDOWS
+    #if MICROPY_NLR_OS_WINDOWS
     "movq   88(%%rcx), %%rsi    \n" // load saved %rsi
     "movq   80(%%rcx), %%rdi    \n" // load saved %rdr
     #endif

py/nlrx86.c

@@ -33,13 +33,7 @@
 // For reference, x86 callee save regs are:
 //  ebx, esi, edi, ebp, esp, eip
 
-#if defined(_WIN32) || defined(__CYGWIN__)
-#define NLR_OS_WINDOWS 1
-#else
-#define NLR_OS_WINDOWS 0
-#endif
-
-#if NLR_OS_WINDOWS
+#if MICROPY_NLR_OS_WINDOWS
 unsigned int nlr_push_tail(nlr_buf_t *nlr) asm("nlr_push_tail");
 #else
 __attribute__((used)) unsigned int nlr_push_tail(nlr_buf_t *nlr);