Start using file_util symlink code, now that it's available.

base/base_paths_linux.cc

Index: base/base_paths_linux.cc
diff --git a/base/base_paths_linux.cc b/base/base_paths_linux.cc
index 701b9830f6d1b5a4c74aed923361d5c2c13d6c54..3246775fc3f991a88e23ac5cddbc5d605dacc1e3 100644
--- a/base/base_paths_linux.cc
+++ b/base/base_paths_linux.cc
@@ -37,14 +37,12 @@ bool PathProviderPosix(int key, FilePath* result) {
     case base::FILE_EXE:
     case base::FILE_MODULE: {  // TODO(evanm): is this correct?
 #if defined(OS_LINUX)
-      char bin_dir[PATH_MAX + 1];
-      int bin_dir_size = readlink(kSelfExe, bin_dir, PATH_MAX);
-      if (bin_dir_size < 0 || bin_dir_size > PATH_MAX) {

[Evan Martin, 2010/11/30 21:09:30]

It seems the new code loses the >PATH_MAX check.  Do we care?

[Evan Martin, 2010/11/30 21:09:30]

It seems the new code loses the >PATH_MAX check.  Do we care?

[Greg Spencer (Chromium), 2010/11/30 22:15:07]

I don't think we do.  If readlink is saying it copied more bytes into the buffer
than the value supplied to it for the size of the buffer, then that's a pretty
serious bug in readlink, I think.

Also, readlink doesn't NULL terminate, so there's no reason to make the buffer
PATH_MAX + 1.

+      FilePath bin_dir;
+      if (!file_util::ReadSymbolicLink(FilePath(kSelfExe), &bin_dir)) {
         NOTREACHED() << "Unable to resolve " << kSelfExe << ".";
         return false;
       }
-      bin_dir[bin_dir_size] = 0;
-      *result = FilePath(bin_dir);
+      *result = bin_dir;
       return true;
 #elif defined(OS_FREEBSD)
       int name[] = { CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1 };