fix: avoid FrozenError in sysread with frozen buffers

[seuros]

This avoids having error when the caller send frozen strings.

[ioquatix]

I'm not sure I understand why this is desirable? If you have this scenario yourself, why not +buffer as the argument?

[seuros]

It happen on the async adapter such as https://github.com/seuros/activecypher/actions/runs/19809656365/job/56749520537

Let me check if i can fix it upstream.

[seuros]

just double checked, this gem freeze its internal StringBuffer instances for zero-copy reads.

The error happen only connection are interrupt or timeout. (Read operation timed out after 15s)

Right now it crash with FrozenError instead of the correct error.

This fixed it.

[samuel-williams-shopify]

I looked at the activecypher test suite and found various issues, e.g.:

      tasks = 5.times.map do |i|
        @connection.async_write_transaction do |tx|
          tx.run('CREATE (n:TestNode {name: $name, value: $value})',
                 { name: "concurrent-#{i}", value: i * 10 })
          i
        end
      end

You are creating 5 transactions all using the same connection / IO and they are potentially all racing on reads/writes. This can result in failures like the one you are seeing. I'll try to capture more details, but it looks like a test suite issue to me.

[samuel-williams-shopify]

If you check this report you can see the root cause of the issue: https://github.com/samuel-williams-shopify/activecypher/actions/runs/19911580352/job/57081281300