Selector.select throws "private method 'select' called for #<NIO::Selector:0x00007ff9b830fca0> (NoMethodError)" about nio4r HOT 5 CLOSED

I haven't seen this recently, and I'm working on a lot of Puma test changes. Or, I've looked at way too many CI logs recently.

One interesting one was a macOS job where the rake compile step failed, as 'build tools' were missing. We do absolutely nothing with build tools, except for ragel.

Regardless, I'll close and open (or re-open) if I see the issue.

from nio4r.

Thanks for the report, I'll investigate.

from nio4r.

I asked @ko1 about it and he mentioned that Kernel#select is a private method.

irb#1(main):001:0> o = Object.new
=> #<Object:0x00007fa45739bb80>
irb#1(main):002:0> o.select(1)
(irb#1):2:in `<top (required)>': private method `select' called for #<Object:0x00007fa45739bb80> (NoMethodError)

Actually I'm a little surprised it's part of Kernel. Anyway, it seems like that's possibly generating the error message. In other words, maybe NIO::Selector#select is being undefined somehow???

from nio4r.

Thanks for checking.

In other words, maybe NIO::Selector#select is being undefined somehow???

I guess I'll see if I can find if it is. I've been thru the Puma code, don't recall anything in the 'lib' code, and things like that I'd normally consider a very bad idea in the test system. I'll look more tomorrow...

from nio4r.

I checked the code, and a single instance of NIO::Selector is created in one file/object Puma::Reactor. It is an internal object, as code outside of the class cannot access it. The following methods are called:

close, deregister, register, select, wakeup

Or, there is no code that would undefine a method. A few notes:

1. Puma runs some tests parallel, which makes it difficult to determine the results of 'crash & burn' issues. It appears that the test that generated the output passed.

2. I've looked at test logs quite a bit, and I don't recall seeing this before.

3. Using the seed value of the failing test run, I ran the test suite in CI and locally (forcing my system's 'Etc.nprocessors' to match CI systems). Of course, the tests did not repo the error.

4. Given the above, not sure if the error is related to the Ruby version used, which was 3.2.

I've added code to rescue NoMethodError, and it seems to not cause any issues with Puma. Need to do some more testing.

from nio4r.