"Data Loss!"

[kengruven]

I was playing with the new web interface and not seeing any action from my fixtures, so I ran ola_dmxmonitor to make sure the data was getting sent to OLA. Every now and then, it pops up a little text alert that says "Data Loss!".

There doesn't seem to be any documentation anywhere which says what data is being lost, or if this is a problem, or what I should do about it. Or is this a bug? Intuitively, I wouldn't think that moving one slider in a web interface, with one terminal window monitoring the state of 500 bytes, should be enough to cause any kind of data loss.

[kengruven]

This warning seems to be generated by

ola/examples/ola-dmxmonitor.cpp

Line 374 in 1a68747

if (diff > TimeInterval(2, 5000000)) {

which is a really weird line.

According to the TimeInterval header (and USEC_IN_SECONDS), it's seconds and microseconds, so this looks like it means 2 seconds + 5 million microseconds = 7 seconds.

But the usec field is stored in type suseconds_t, which according to POSIX is only required to hold values in the range [-1, 1000000], and I can't find any documentation on what it might do when given a value outside of this range. (Wrap? Ignore? Overflow into the seconds field?)

My computer is pretty new, has lots of idle cores, plenty of RAM, and an SSD, and I can see the DMX monitor continuously update. I doubt it's ever gone even 1 second without a screen update. Whatever this code was intended to do, I'm skeptical it's actually doing that.

[kengruven]

It looks like a lot of the functionality (TimerAdd, TimerSub, Set, etc) assumes that the microsecond count is in the range [0,999999]. That's a perfectly reasonable invariant, but it's never checked or enforced.

I'd propose:

1. Update TimeInterval (or its base class) to check/enforce the range of its inputs
2. Fix ola-dmxmonitor.cpp to not pass an out-of-range value here: I assume the intent was USEC_IN_SECONDS / 2?
3. Document what "Data Loss!" means here. Or rename it: it seems to mean "dropped frame", which is common terminology in many other contexts

It looks like the current line (using glibc) is probably equivalent to:

if (diff > TimeInterval(3, 0)) {

but I'm still skeptical that a terminal window constantly displaying updates of 1/2 KB of data is ever exceeding 3sec of lag.