Warn on zero-bufcnt target (based on v87pre3)

[ecashin]

No description provided.

[mischief]

i believe that the code will fail to issue the warning if ch->bufcnt (n) is 1, because of t->maxout = t->nframes / 2;, so t->maxout will still be 0 when 1 is divided by 2.

so, maybe check for 1 as well?

[ecashin]

How's that?

The t->maxout is really a misnomer that should be cwnd in deference to Van Jacobson's description of TCP congestion control. The current size of the dynamic estimation of the available bandwidth, so it gets reset by aoecmd_wreset at times, becoming one. We just need to leave it there if nframes is one.

It would be worth proving that a bufcnt==one target can initialize and get used without trouble, though. If you have time to try that on this branch, I'd love to hear about your results.

[ecashin]

@mischief, hi. Did you notice this branch? It would be great to hear how these changes work for your misbehaving target.