debing.sun

It feels like you are missing something, `XGROUPREAD`is mutually exclusive, if message A is consumed by consumer c1, c2 will not receive message A. If we use push, we would...

Normally, the network latency would be much less than the time consumer consume message, so why do you need such a low latency?

@addos on different CPU architecture or platform, floating precision will be different, and this test had been skipped on non-x86 platform in https://github.com/redis/redis/commit/fceaa46dda37a2d4db4e0a48cc1dab3fe208cb2c.

@addos can you use `redis-cli info server` to see what version of redis you are using?

@addos Can you see what `uname -a` outputs? This test is skipped in my mac m1.

@addos It's odd that this test never enters under mac m1. can you try `./runtest --single unit/type/hash --only "Test HINCRBYFLOAT for correct float representation*"` again and see the output?

@addos thanks, but it's so wired, `Test HINCRBYFLOAT for correct float representation` was skipped, normally, `make test` will also skip it, do you have any special steps?

@jgaskins Because length of `my-stream`' is 0, `XREADGROUP` command will block until times out, during this time the consumer is not actually read, because there is no need to read...

@jgaskins It does seem to be ambiguous. Just like when `XREADGROUP` returns an error, the consumer's idle is not updated. ping @itamarhaber

Same as #11069