Make SSO support embedded \0s like native String

Supercedes esp8266#6027

Make SSO more generic by keeping track of its length explicitly,
allowing for embedded \0s to exist in the String (just like the non-SSO
ones).

Author: earlephilhower

cores/esp8266/WString.cpp

@@ -136,7 +136,7 @@ inline void String::init(void) {
 }
 
 void String::invalidate(void) {
-    if(!sso() && wbuffer())
+    if(!isSSO() && wbuffer())
         free(wbuffer());
     init();
 }
@@ -154,17 +154,19 @@ unsigned char String::reserve(unsigned int size) {
 
 unsigned char String::changeBuffer(unsigned int maxStrLen) {
     // Can we use SSO here to avoid allocation?
-    if (maxStrLen < sizeof(sso_buf)) {
-        if (sso() || !buffer()) {
+    if (maxStrLen < sizeof(sso.buff) - 1) {
+        if (isSSO() || !buffer()) {
             // Already using SSO, nothing to do
             setSSO(true);
             return 1;
-        } else { // if bufptr && !sso()
-            // Using bufptr, need to shrink into sso_buff
-            char temp[sizeof(sso_buf)];
+        } else { // if bufptr && !isSSO()
+            // Using bufptr, need to shrink into sso.buff
+            char temp[sizeof(sso.buff)];
             memcpy(temp, buffer(), maxStrLen);
             free(wbuffer());
+            uint16_t oldLen = len();
             setSSO(true);
+	    setLen(oldLen);
             memcpy(wbuffer(), temp, maxStrLen);
             return 1;
         }
@@ -176,12 +178,12 @@ unsigned char String::changeBuffer(unsigned int maxStrLen) {
         return false;
     }
     uint16_t oldLen = len();
-    char *newbuffer = (char *) realloc(sso() ? nullptr : wbuffer(), newSize);
-    if(newbuffer) {
+    char *newbuffer = (char *) realloc(isSSO() ? nullptr : wbuffer(), newSize);
+    if (newbuffer) {
         size_t oldSize = capacity() + 1; // include NULL.
-        if (sso()) {
+        if (isSSO()) {
             // Copy the SSO buffer into allocated space
-            memcpy(newbuffer, sso_buf, sizeof(sso_buf));
+            memcpy(newbuffer, sso.buff, sizeof(sso.buff));
         }
         if (newSize > oldSize)
         {
@@ -229,15 +231,15 @@ void String::move(String &rhs) {
 	    rhs.invalidate();
             return;
         } else {
-            if (!sso()) {
+            if (!isSSO()) {
                 free(wbuffer());
                 setBuffer(nullptr);
             }
         }
     }
-    if (rhs.sso()) {
+    if (rhs.isSSO()) {
         setSSO(true);
-        memmove(sso_buf, rhs.sso_buf, sizeof(sso_buf));
+        memmove(sso.buff, rhs.sso.buff, sizeof(sso.buff));
     } else {
         setSSO(false);
         setBuffer(rhs.wbuffer());

cores/esp8266/WString.h

@@ -250,27 +250,29 @@ class String {
             uint16_t cap;
             uint16_t len;
         };
-
-        // SSO is handled by checking the last byte of sso_buff.
-        // When not in SSO mode, that byte is set to 0xff, while when in SSO mode it is always 0x00 (so it can serve as the string terminator as well as a flag)
-        // This allows strings up up to 12 (11 + \0 termination) without any extra space.
-        enum { SSOSIZE = sizeof(struct _ptr) + 4 }; // Characters to allocate space for SSO, must be 12 or more
+        // This allows strings up up to 11 (10 + \0 termination) without any extra space.
+        enum { SSOSIZE = sizeof(struct _ptr) + 4 - 1 }; // Characters to allocate space for SSO, must be 12 or more
+        struct _sso {
+            char     buff[SSOSIZE];
+            unsigned len  : 7;
+            unsigned flag : 1;
+        };
         enum { CAPACITY_MAX = 65535 }; // If size of capacity changed, be sure to update this enum
         union {
             struct _ptr ptr;
-            char sso_buf[SSOSIZE];
+            struct _sso sso;
         };
         // Accessor functions
-        inline bool sso() const { return sso_buf[SSOSIZE - 1] == 0; }
-        inline unsigned int len() const { return sso() ? strlen(sso_buf) : ptr.len; }
-        inline unsigned int capacity() const { return sso() ? SSOSIZE - 1 : ptr.cap; }
-        inline void setSSO(bool sso) { sso_buf[SSOSIZE - 1] = sso ? 0x00 : 0xff; }
-        inline void setLen(int len) { if (!sso()) ptr.len = len; }
-        inline void setCapacity(int cap) { if (!sso()) ptr.cap = cap; }
-	inline void setBuffer(char *buff) { if (!sso()) ptr.buff = buff; }
+        inline bool isSSO() const { return sso.flag; }
+        inline unsigned int len() const { return isSSO() ? sso.len : ptr.len; }
+        inline unsigned int capacity() const { return isSSO() ? (unsigned int)SSOSIZE - 1 : ptr.cap; }
+        inline void setSSO(bool set) { sso.flag = set; }
+        inline void setLen(int len) { if (!isSSO()) ptr.len = len; else sso.len = len; }
+        inline void setCapacity(int cap) { if (!isSSO()) ptr.cap = cap; }
+	inline void setBuffer(char *buff) { if (!isSSO()) ptr.buff = buff; }
         // Buffer accessor functions
-        inline const char *buffer() const { return (const char *)(sso() ? sso_buf : ptr.buff); }
-        inline char *wbuffer() const { return sso() ? const_cast<char *>(sso_buf) : ptr.buff; } // Writable version of buffer
+        inline const char *buffer() const { return (const char *)(isSSO() ? sso.buff : ptr.buff); }
+        inline char *wbuffer() const { return isSSO() ? const_cast<char *>(sso.buff) : ptr.buff; } // Writable version of buffer
 
     protected:
         void init(void);

tests/host/core/test_string.cpp

@@ -320,11 +320,11 @@ TEST_CASE("String SSO works", "[core][String]")
   REQUIRE(s.c_str() == savesso);
   REQUIRE(s == "0123456789");
   REQUIRE(s.length() == 10);
-  s += "a";
-  REQUIRE(s.c_str() == savesso);
-  REQUIRE(s == "0123456789a");
-  REQUIRE(s.length() == 11);
   if (sizeof(savesso) == 4) {
+    s += "a";
+    REQUIRE(s.c_str() != savesso);
+    REQUIRE(s == "0123456789a");
+    REQUIRE(s.length() == 11);
     s += "b";
     REQUIRE(s.c_str() != savesso);
     REQUIRE(s == "0123456789ab");
@@ -334,12 +334,16 @@ TEST_CASE("String SSO works", "[core][String]")
     REQUIRE(s == "0123456789abc");
     REQUIRE(s.length() == 13);
   } else {
+  s += "a";
+    REQUIRE(s.c_str() == savesso);
+    REQUIRE(s == "0123456789a");
+    REQUIRE(s.length() == 11);
     s += "bcde";
     REQUIRE(s.c_str() == savesso);
     REQUIRE(s == "0123456789abcde");
     REQUIRE(s.length() == 15);
     s += "fghi";
-    REQUIRE(s.c_str() == savesso);
+    REQUIRE(s.c_str() != savesso);
     REQUIRE(s == "0123456789abcdefghi");
     REQUIRE(s.length() == 19);
     s += "j";
@@ -360,7 +364,7 @@ TEST_CASE("String SSO works", "[core][String]")
     REQUIRE(s.length() == 23);
     s += "nopq";
     REQUIRE(s.c_str() != savesso);
-      REQUIRE(s == "0123456789abcdefghijklmnopq");
+    REQUIRE(s == "0123456789abcdefghijklmnopq");
     REQUIRE(s.length() == 27);
     s += "rstu";
     REQUIRE(s.c_str() != savesso);