On 03/23/2017 05:38 AM, Howard Chu wrote:

subbarao@computer.org wrote:

...

This could well be a very good idea for broader reasons, but it probably would be less desirable as a sole solution for this particular situation, at least for my use case. In this case, I'm backporting 2.4.44 for zesty to 14.04 LTS. Ideally, I want this to be a clean rebuild, without any code changes.

Overriding shell variables doesn't require any code changes. It only requires that you set their values in whatever build script you use that issues the "make test" command. That is the whole reason why we defined these variables.

Hmm, I don't think you're hearing what I'm saying in its intended context. But let me table this for now. Can we back up to the other part of the thread? If you can point out what I'm missing with the VM clock reliability, that would help me better appreciate where you're coming from.

For example, if we were talking about tests between multiple VMs, I could immediately appreciate the value of being generous with timing margins. But as I indicated in my other message, I'm having difficulty seeing how those kinds of problems could arise here.

I apologize if there's something obvious I'm missing or if I'm coming across as belaboring an undesirable point. I'm just trying to make sure we don't inadvertently rule out a simple solution to the problem that could work for everyone.

Thanks,

-Kartik