| || |
Debugging Linux LoadGenerator problems
I'm new to LR. I'm using LR 11.0 and trying to debug a problem with a couple of shared libraries I load from a LR script on a Linux LoadGenerator (RHEL5). I have not been able to find any documentation about what the execution environment is for the multi-threaded mdrv on Linux so the development process has been trial and error so far.
These shared libraries work fine from my unit test drivers in single thread and multi-thread configurations, but fail when invoked by mdrv. The mdrv log file reports a memory violation, but I guess mdrv has a signal handler and doesn't produce a core file to work with.
So, I guess my questiosn are these:
Is there any documentation about writing/compiling shared libraries for LR Linux LoadGenerator or examples?
Is there a way to run/debug mdrv interactively or have it produce a core file on Linux?
One problem seems to be related to varargs processing in the shared libraries, I compile/link with -m32 -fPIC -shared, which works fine for the unit tests, but seems to fail in mdrv. Sound familiar?
Thanks in advance.
Re: Debugging Linux LoadGenerator problems
I can't think of a way to force mdrv on linux to do a core dump, but I wonder if some off the interpreter options force a higher level of tracing could be useful. In particular I was thinking of -cci_debug, -cci_trace and/or -cci_elevel, to get some more information on what's been happening in your environment.
I think HP tier 3 support (typically dev in Israel) may have some additional insight on affecting a core dump, but getting to them through standard support takes forever.
Replace ineffective offshore contracts, LoadRunnerByTheHour
. Starting @ $19.95/hr USD.
Put us to the test, skilled expertise is less expensive than you might imagine.
Twitter: @LoadRunnerBTH @PerfBytes