While running a session using WWW scripts, I sometimes get a QALoad error that states that my VUser crashed in the middleware. I look at the .rip file and nothing appears wrong. Does anyone know a way to make this generic message more specific or a way to dig a little deeper to find out what is happening?
I have run into this issue when QALoad is looking for an array when a constant is declared or vice versa, especially when buffers are being captured. Once I changed the variable type, the middleware crash stopped 99% of the time.
I haven't mad any mods to the scripts. I can run the tests against one version of the software and never receive the middleware error. I then run the same scripts against a later version of the software and I get the errors occasionally. Functionally, the application checks out. I can't find any perfmon stats that point to anything. When I do receive the error, it isn't repeatable with a specific Vuser.
We too get Middleware exception errors during the test, and all the users get aborted, What we did is "we just uninstalled the Windows XP and installed the Windows 2000 Profession, and now its working fine without any problem. [img]images/icons/cool.gif[/img]
Sounds as though there is a coding error (or the app returning unexpected data) causing the VU thread to crash, not exit out as it should. When a crash occurs, usually due to some coding issue like a buffer overrun, it causes the thread to die without writing any cause to the RIP file.
The solution as per the last post is somewhat extreme - I wouldn't necessarily go this far an reinstall an older OS as QAL will happily exist under XP.
Trick is to try and isolate which call is causing the problem. If it is some code you have added you may want to ensure that you've allocated memory appropriately.
Not to put the blame all on you (!), it could also be the case that the error is caused in a QALoad call where it could be the app under test returning something we were not expecting. If this is the case, it is reproducable and you can tell us what command is causing this, please log this with support as there may be something environmental that we need to support.