Decode bytes when formatting them as strings

USDTProbe objects (and USDTProbeArguments and USDTProbeLocations) are
instantiated with data that's sourced from libccc calls. That means
that their attributes are bytes-typed, not string-typed.

When a bytes-typed value is rendered into a string with Python's '%s'
formatting directive, it gets a wrapped in single quotes and prefixed
with b. For example, b'probe-location'. This is visually noisy, but
also breaks some tool behavior which uses string-formatted values for
stuff like filters.

This is only an issue in Python 3. In Python 2, the bytes type is just
an alias for the string type, and so byte sequences from libcc were
implicitly decoded as ASCII text.

diff --git a/src/python/bcc/usdt.py b/src/python/bcc/usdt.py
index 87f68d111ad3623626a0ee6455efdd357829d2c8..b96ac0eb44802847c6a3bd4145b0bb4ef64751cc 100644 (file)
--- a/src/python/bcc/usdt.py
+++ b/src/python/bcc/usdt.py
@@ -50,24 +50,24 @@ class USDTProbeArgument(object):
         if self.valid & BCC_USDT_ARGUMENT_FLAGS.CONSTANT != 0:
             return "%d" % self.constant
         if self.valid & BCC_USDT_ARGUMENT_FLAGS.DEREF_OFFSET == 0:
-            return "%s" % self.base_register_name
+            return "%s" % self.base_register_name.decode()
         if self.valid & BCC_USDT_ARGUMENT_FLAGS.DEREF_OFFSET != 0 and \
            self.valid & BCC_USDT_ARGUMENT_FLAGS.DEREF_IDENT == 0:
             if self.valid & BCC_USDT_ARGUMENT_FLAGS.INDEX_REGISTER_NAME != 0:
-                index_offset = " + %s" % self.index_register_name
+                index_offset = " + %s" % self.index_register_name.decode()
                 if self.valid & BCC_USDT_ARGUMENT_FLAGS.SCALE != 0:
                     index_offset += " * %d" % self.scale
             else:
                 index_offset = ""
             sign = '+' if self.deref_offset >= 0 else '-'
-            return "*(%s %s %d%s)" % (self.base_register_name,
-                                    sign, abs(self.deref_offset), index_offset)
+            return "*(%s %s %d%s)" % (self.base_register_name.decode(),
+                                      sign, abs(self.deref_offset), index_offset)
         if self.valid & BCC_USDT_ARGUMENT_FLAGS.DEREF_OFFSET != 0 and \
            self.valid & BCC_USDT_ARGUMENT_FLAGS.DEREF_IDENT != 0 and \
            self.valid & BCC_USDT_ARGUMENT_FLAGS.BASE_REGISTER_NAME != 0 and \
            self.base_register_name == "ip":
             sign = '+' if self.deref_offset >= 0 else '-'
-            return "*(&%s %s %d)" % (self.deref_ident,
+            return "*(&%s %s %d)" % (self.deref_ident.decode(),
                                      sign, abs(self.deref_offset))
         # If we got here, this is an unrecognized case. Doesn't mean it's
         # necessarily bad, so just provide the raw data. It just means that
@@ -86,7 +86,7 @@ class USDTProbeLocation(object):
         self.bin_path = location.bin_path
 
     def __str__(self):
-        return "%s 0x%x" % (self.bin_path, self.address)
+        return "%s 0x%x" % (self.bin_path.decode(), self.address)
 
     def get_argument(self, index):
         arg = bcc_usdt_argument()
@@ -111,10 +111,10 @@ class USDTProbe(object):
 
     def __str__(self):
         return "%s:%s [sema 0x%x]" % \
-               (self.provider, self.name, self.semaphore)
+               (self.provider.decode(), self.name.decode(), self.semaphore)
 
     def short_name(self):
-        return "%s:%s" % (self.provider, self.name)
+        return "%s:%s" % (self.provider.decode(), self.name.decode())
 
     def get_location(self, index):
         loc = bcc_usdt_location()

diff --git a/tools/tplist.py b/tools/tplist.py
index 6ec2fbe18b92dc444022e1cd9a097a6a9b2422b8..24e67d60b924ebb93c0f32f0b12b1c448c82b51c 100755 (executable)
--- a/tools/tplist.py
+++ b/tools/tplist.py
@@ -80,8 +80,8 @@ def print_usdt_details(probe):
                         print("  %d location(s)" % probe.num_locations)
                         print("  %d argument(s)" % probe.num_arguments)
         else:
-                print("%s %s:%s" %
-                      (probe.bin_path, probe.provider, probe.name))
+                print("%s %s" %
+                      (probe.bin_path.decode(), probe.short_name()))
 
 def print_usdt(pid, lib):
         reader = USDT(path=lib, pid=pid)