llvm-cov: When reading strings in gcov data, skip leading zeros
authorJustin Bogner <mail@justinbogner.com>
Thu, 27 Mar 2014 00:06:36 +0000 (00:06 +0000)
committerJustin Bogner <mail@justinbogner.com>
Thu, 27 Mar 2014 00:06:36 +0000 (00:06 +0000)
It seems that gcov, when faced with a string that is apparently zero
length, just keeps reading words until it finds a length it likes
better. I'm not really sure why this is, but it's simple enough to
make llvm-cov follow suit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@204881 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/Support/GCOV.h

index aeac4555cee05b29e49748eff54952bd47ccbf7d..902f2dbc83bf018dba3db05d26ae862235a13bb8 100644 (file)
@@ -204,8 +204,11 @@ public:
   }
 
   bool readString(StringRef &Str) {
-    uint32_t Len;
-    if (!readInt(Len)) return false;
+    uint32_t Len = 0;
+    // Keep reading until we find a non-zero length. This emulates gcov's
+    // behaviour, which appears to do the same.
+    while (Len == 0)
+      if (!readInt(Len)) return false;
     Len *= 4;
     if (Buffer->getBuffer().size() < Cursor+Len) {
       errs() << "Unexpected end of memory buffer: " << Cursor+Len << ".\n";