I started using this miner and I like it. I have a big issue with it though: many times the error message does not state which GPU crashed, it just says "Unrecoverable errors" without identifying which GPU. DSTM is better in that regard, as it always states which GPU and hence allows me to fine tune overclocking in rigs with many GPUs. Could you please add clear GPU ids in the error messages to allow us to parse logs and automate other functions?
Also, with bminer I get errors that do not produce a Xid error in the kernel -- this is weird and it's the first time it happens. All other miners I used (for various coins) trigger a Xid error. It sounds like some crashes are in fact bminer's fault rather than the hardware or too much overclocking. I have no way of debugging it though being closed source.
I think the GPU id is available in the error message. The error message looks like "Irrecoverable errors from miner in GPU 0 ..."
The Xid error I have ever seen is due to hardware errors. Can you paste the error logs here?
bminer is tested extensively under Linux and I'm not aware of Xid errors that are caused by the miner itself. However, bminer does stretch the GPU a little bit more for even more hashrate, so if you overclock the GPU you might want to tune the parameters a little bit.
No, you do not specify the GPU id in all error messages. Here's an example:
[FATA] [2018-01-03T16:05:03Z] Fatal cuda error. Terminate soon...
[WARN] [2018-01-03T16:05:08Z] Miner died! It will be restarted soon...
Just search your source code for error messages and you'll see which ones do not specify the GPU id.
Also, you misunderstood my point about Xid kernel driver errors. Once again, some of the errors reported by bminer are NOT also accompanied by a Xid driver error (in kern.log). No other miner does this. It's a problem when bminer doesn't report the GPU id in the error message since I have no idea which GPU is at fault (no Xid, no GPU id in bminer error).
Also, if the miner reports a problem with the hardware but the nvidia driver sees nothing wrong, then it's likely that you have a bug in your code. No other miner has this behavior - I tested dstm, ewbf, ethminer, claymore, ccminer for a long time; all errors reported by these miners are accompanied by Xid errors - that is, a hardware fault did indeed occur.
I don't think bminer is tested as extensively as you claim - it can't, really, given how new it is.
My system is a Ubuntu 16.04, nvidia 384.90, 384.98, 387.34, kernel 4.4.0.104.
I like bminer's speed though, which is why I'm taking the time to post here to help you.