Merge pull request #1 from dscho/git-for-windows

Assorted fixes for Git for windows

Author: dscho

newlib/libc/stdlib/mbtowc_r.c

@@ -36,7 +36,7 @@ __ascii_mbtowc (struct _reent *r,
   if (n == 0)
     return -2;
 
-#ifdef __CYGWIN__
+#ifdef STRICTLY_7BIT_ASCII
   if ((wchar_t)*t >= 0x80)
     {
       r->_errno = EILSEQ;

newlib/libc/stdlib/wctomb_r.c

@@ -29,7 +29,7 @@ __ascii_wctomb (struct _reent *r,
   if (s == NULL)
     return 0;
 
-#ifdef __CYGWIN__
+#ifdef STRICTLY_7BIT_ASCII
   if ((size_t)wchar >= 0x80)
 #else
   if ((size_t)wchar >= 0x100)

winsup/cygwin/environ.cc

@@ -1192,7 +1192,15 @@ build_env (const char * const *envp, PWCHAR &envblock, int &envc,
       {
 	/* Don't pass timezone environment to non-msys applications */
 	if (ascii_strncasematch(*srcp, "TZ=", 3))
-	  goto next1;
+          {
+	    const char *v = *srcp + 3;
+	    if (*v == ':')
+	      goto next1;
+	    for (; *v; v++)
+	      if (!isalpha(*v) && !isdigit(*v) &&
+		  *v != '-' && *v != '+' && *v != ':')
+	        goto next1;
+	  }
 	else if (ascii_strncasematch(*srcp, "MSYS2_ENV_CONV_EXCL=", 20))
 	  {
 	    msys2_env_conv_excl_env = (char*)alloca (strlen(&(*srcp)[20])+1);
@@ -1304,7 +1312,7 @@ build_env (const char * const *envp, PWCHAR &envblock, int &envc,
 	     Note that this doesn't stop invalid strings without '=' in it
 	     etc., but we're opting for speed here for now.  Adding complete
 	     checking would be pretty expensive. */
-	  if (len == 1 || !*rest)
+	  if (len == 1)
 	    continue;
 
 	  /* See if this entry requires posix->win32 conversion. */

winsup/cygwin/how-to-debug-cygwin.txt

@@ -126,3 +126,9 @@ set CYGWIN_DEBUG=cat.exe:gdb.exe
    program will crash, probably in small_printf.  At that point, a 'bt'
    command should show you the offending call to strace_printf with the
    improper format string.
+
+9. Debug output without strace
+
+   If you cannot use gdb, or if the program behaves differently using strace
+   for whatever reason, you can still use the small_printf() function to
+   output debugging messages directly to stderr.

winsup/cygwin/msys2_path_conv.cc

@@ -200,7 +200,7 @@ void ppl_convert(const char** from, const char* to, char** dst, const char* dste
 
 
 void find_end_of_posix_list(const char** to, int* in_string) {
-    for (; **to != '\0' && (in_string ? (**to != *in_string) : **to != ' '); ++*to) {
+    for (; **to != '\0' && (!in_string || **to != *in_string); ++*to) {
     }
 
     if (**to == *in_string) {
@@ -237,6 +237,7 @@ void find_end_of_rooted_path(const char** from, const char** to, int* in_string)
 void sub_convert(const char** from, const char** to, char** dst, const char* dstend, int* in_string) {
     const char* copy_from = *from;
     path_type type = find_path_start_and_type(from, false, *to);
+    debug_printf("found type %d for path %s", type, copy_from);
 
     if (type == POSIX_PATH_LIST) {
         find_end_of_posix_list(to, in_string);
@@ -300,12 +301,6 @@ const char* convert(char *dst, size_t dstlen, const char *src) {
             }
             continue;
         }
-
-        if (isspace(*srcit)) {
-            //sub_convert(&srcbeg, &srcit, &dstit, dstend, &in_string);
-            //srcbeg = srcit + 1;
-            break;
-        }
     }
 
     sub_convert(&srcbeg, &srcit, &dstit, dstend, &in_string);
@@ -345,8 +340,67 @@ path_type find_path_start_and_type(const char** src, int recurse, const char* en
 
     if (*it == '\0' || it == end) return NONE;
 
-    if (!isalnum(*it) && *it != '/' && *it != '\\' && *it != ':' && *it != '-' && *it != '.') {
-        return find_path_start_and_type(move(src, 1), true, end);
+    /* Let's not convert ~/.file to ~C:\msys64\.file */
+    if (*it == '~') {
+skip_p2w:
+        *src = end;
+        return NONE;
+    }
+
+    /*
+     * Skip path mangling when environment indicates it.
+     */
+    const char *no_pathconv = getenv ("MSYS_NO_PATHCONV");
+
+    if (no_pathconv)
+      goto skip_p2w;
+
+    /*
+     * Prevent Git's :file.txt and :/message syntax from beeing modified.
+     */
+    if (*it == ':')
+        goto skip_p2w;
+
+    while (it != end && *it) {
+        switch (*it) {
+        case '`':
+        case '\'':
+        case '"':
+        case '*':
+        case '?':
+        case '[':
+        case ']':
+            goto skip_p2w;
+        case '/':
+            if (it + 1 < end && it[1] == '~')
+                goto skip_p2w;
+            break;
+        case ':':
+            // Avoid mangling IPv6 addresses
+            if (it + 1 < end && it[1] == ':')
+                goto skip_p2w;
+
+            // Leave Git's <rev>:./name syntax alone
+            if (it + 1 < end && it[1] == '.') {
+                if (it + 2 < end && it[2] == '/')
+                    goto skip_p2w;
+                if (it + 3 < end && it[2] == '.' && it[3] == '/')
+                    goto skip_p2w;
+            }
+            break;
+        case '@':
+            // Paths do not contain '@@'
+            if (it + 1 < end && it[1] == '@')
+                goto skip_p2w;
+        }
+        ++it;
+    }
+    it = *src;
+
+    while (!isalnum(*it) && !(0x80 & *it) && *it != '/' && *it != '\\' && *it != ':' && *it != '-' && *it != '.') {
+        recurse = true;
+        it = ++*src;
+        if (it == end || *it == '\0') return NONE;
     }
 
     path_type result = NONE;
@@ -408,6 +462,8 @@ path_type find_path_start_and_type(const char** src, int recurse, const char* en
 
     int starts_with_minus = 0;
     int starts_with_minus_alpha = 0;
+    int only_dots = *it == '.';
+    int has_slashes = 0;
     if (*it == '-') {
       starts_with_minus = 1;
       it += 1;
@@ -451,11 +507,17 @@ path_type find_path_start_and_type(const char** src, int recurse, const char* en
                 if (ch == '/' && *(it2 + 1) == '/') {
                     return URL;
                 } else {
+                    if (!only_dots && !has_slashes)
+                        goto skip_p2w;
                     return POSIX_PATH_LIST;
                 }
             } else if (memchr(it2, '=', end - it) == NULL) {
                 return SIMPLE_WINDOWS_PATH;
             }
+        } else if (ch != '.') {
+            only_dots = 0;
+            if (ch == '/' || ch == '\\')
+                has_slashes = 1;
         }
     }
 

winsup/cygwin/strfuncs.cc

@@ -641,7 +641,11 @@ sys_cp_mbstowcs (mbtowc_p f_mbtowc, wchar_t *dst, size_t dlen,
 	     to store them in a symmetric way. */
 	  bytes = 1;
 	  if (dst)
+#ifdef STRICTLY_7BIT_ASCII
 	    *ptr = L'\xf000' | *pmbs;
+#else
+	    *ptr = *pmbs;
+#endif
 	  memset (&ps, 0, sizeof ps);
 	}