rec: Sanitize d_orig_ttl stored in record cache

[omoerbeek]

The computed d_orig_ttl can be wrong if d_now.tv_sec is more recent than the d_now.tv_sec the TTD was computed from. This can happen if we went out to get e.g. keys and that took time. d_orig_ttl even potentially wraps around in that case if the TTL value was smaller than that delay. Work around that and make sure d_orig_ttl is within the legal range.

An alternative approach would be to store d_orig_ttl the moment we change the d_ttl to be a TTD. I looked into that but we have no convenient place to store it. There are also a few places the TTD is replaced yet again with another value, so things started to get quite complicated. But maybe I missed an approach to have the real original TTL stored somewhere. I'll keep looking into that, as that is a better approach in the end. Hence the draft status.

Note: 2nd commit takes the alternate approach. I jave left the original approach in between #if 0/#endif markers. Those blocks will be removed if the 2nd approach is deemed OK.

Short description

Checklist

I have:

• read the CONTRIBUTING.md document
• compiled this code
• tested this code
• included documentation (including possible behaviour changes)
• documented the code
• added or modified regression test(s)
• added or modified unit test(s)
• checked that this code was merged to master

[Habbie]

I prefer reading the first version. I think I functionally prefer the second version.

Approving this, while leaving you free to choose. Either way looks good.

[rgacogne]

I agree with Peter, both options look fine. The second one is more accurate but also more complicated, I wonder if it would not be easier to record the time when we computed the TTD instead of the original TTL, so we would in theory not need to "compensate" the value later?

[omoerbeek]

I agree with Peter, both options look fine. The second one is more accurate but also more complicated, I wonder if it would not be easier to record the time when we computed the TTD instead of the original TTL, so we would in theory not need to "compensate" the value later?

That looks like a viable thing to investigate, will do so.