Conversion on GET fails with MQRC_CONVERTED_STRING_TOO_BIG when using special characters on machines with different CCSIDs

[Cottonglow]

In the following scenario:

1. Create a queue manager on a machine that is using CCSID 819
2. Update the queue manager CCSID using the following ALTER QMGR CCSID(850)
3. Create a user called Müller - no need to set permissions for this user at this point
4. Run DISPLAY QMGR as Müller - this should fail, causing the name to appear in our error logs
5. On a second machine that is using UTF-8, run the following PCF command MQCMD_INQUIRE_ERROR_LOG to retrieve the error logs.
6. The command then returns the following error:

File "/results/interops/46836/python_packages/ibmmq/mqadmin.py", line 354, in __call__
    raise e
File "/results/interops/46836/python_packages/ibmmq/mqadmin.py", line 303, in __call__
    message = self.__pcf.reply_queue.get(None, get_md, get_opts)
File "/results/interops/46836/python_packages/ibmmq/mqqueue.py", line 298, in get
    raise MQMIError(rv[-2], rv[-1])
mqerrors.MQMIError: MQI Error. Comp 1, Reason 2190: WARNING: MQRC_CONVERTED_STRING_TOO_BIG

I'm not entirely sure if step 2 is necessary to replicate this error message but I have added in my full scenario just in case.

[ibmmqmet]

Thanks for this.

As discussed, data conversion can always expand messages beyond the "original length". And that's inherent in the MQI - even though it's not ideal. Add to that, the fact that the Python layer has always attempted to hide some of the MQI complexities. In this case, by not requiring apps to give a buffer size. But it doesn't help if conversion needs extra space.

I think there are two things that need to be done, that will maintain compatibility without forcing apps to deal with retries:

• Add more headroom when doing the refetch in get() for apps that have not given an explicit max_length
• Allowing PCF operations to also specify a max_length parameter, in case that headroom is still not sufficient.

I'll sort out something for the next update.