Provide a way of logging extra bytes in message

[EmilioPeJu]

It has been encountered some cases in which a reply is padded with extra zero bytes. The reply content is still valid but the extra bytes can be ignored. Logging that condition (instead of raising an exception) will allow returning the reply.

[EmilioPeJu]

Actually, I found that the padding was added by ipmitool, when sending a command to delete a SEL entry, the answer contained an extra zero byte. The ipmitool version used is 1.8.18, not sure if a newer version will not have that issue.

[hthiery]

Are you sure that the extra byte is added by ipmitool? I suspect that the addressed ipmi device sends this additional byte in the response.

[EmilioPeJu]

I reported the issue to the device vendor and they replied saying that they did some test and confirmed that the IPMI reply didn't have an extra zero and therefore it would be ipmitools' doing. I have done a capture using wireshark and the reply is as follows: 81 2c 53 20 28 46 00 00 00 00 72 It looks like it's actually coming from the device. Anyways, I think python-ipmi should have a way of working around these cases.

[EmilioPeJu]

Here is an alternative solution:

diff --git a/pyipmi/msgs/message.py b/pyipmi/msgs/message.py
index c399778..a0bc398 100644
--- a/pyipmi/msgs/message.py
+++ b/pyipmi/msgs/message.py
@@ -385,7 +385,11 @@ class Message(object):
                 break

         if (cc is None or cc == 0) and len(data) > 0:
-            raise DecodingError('Data has extra bytes')
+            self.on_extra_data()
+
+    def on_extra_data(self):
+        """What to do when not expected extra data is found in a message"""
+        raise DecodingError('Data has extra bytes')

     def _is_request(self):
         return self.__netfn__ & 1 == 0

That way, I can monkey patch on_extra_data to not raise an exception (in the code using python-ipmi)

[hthiery]

What about introducing a new Exception type (e.g. MessageExtraDataError)? With that you could catch that and do what you want to do without monkey patching.

[EmilioPeJu]

There are 3 problems with raising an exception:

• The upper calling function (in this case send_message_with_name()) will not return the reply
• Any processing done after the exception is not done anymore.
• My code can do nothing about that exception, given that delete_sel_entry() is called by get_and_clear_sel_entry(), so I would be forced to monkey patch that function to handle the exception.

I need get_and_clear_sel_entry() to work normally in spite of having an extra zero byte in the reply for delete_sel_entry()

[hthiery]

ah I see. sou let's do it the way you proposed.

[EmilioPeJu]

I guess I'm the only one having this issue (and it's not worth making a change addressing it), I can still monkey patch the _decode() function as a workaround

[hthiery]

sorry ... but i had accidentally closed #107. actually i wanted to merge this one.