>>>>> "Nathan" == Nathan Neulinger <Neulinger> writes:
Nathan> You're going to have to change it in several places... the Nathan> makefiles are not quite yet consistent. Nathan> I don't remember every place you have to do it. Are you Nathan> concerned about purely user space code, or kernel module as Nathan> well? Primarily use space code, since fileserver and volserver crashes are what we are most keen to get maximum information for. My experience suggests that running binaries compiled with both -O and -g has minimal costs, and that cost is usually nothing more the extra disk space for the larger binaries, and slightly bigger memory footprint for the processes. The performance overhead has *usually* been minimal. If the same holds true for kernel modules (not my strength), then I'd prefer to build them for maximum debugging information, too. If a debug kernel module costs me significantly, performance wise, then I think I would want to have two builds -- optimized and debug -- and I would configure problematic hosts with the debug version, on a case by case basis, and default to the optimized release. _______________________________________________ OpenAFS-devel mailing list [EMAIL PROTECTED] https://lists.openafs.org/mailman/listinfo/openafs-devel
