THRIFT-6144: Report EOF when Ruby transport reads make no progress

[kpumuk]

Ruby transports may report end of input by returning nil or an empty string. BaseTransport#read_all currently leaks a Ruby NoMethodError for nil, while an empty result causes it to retry forever without making progress.

This change treats either result as TransportException::END_OF_FILE when bytes are still required. Partial reads continue normally while each call makes progress, and zero-size reads still return an empty binary string without invoking the underlying transport.

The common path returns a complete first chunk directly. A short frozen chunk is copied only when it must become the accumulator for subsequent reads.

Benchmarks

The common exact-read path was measured in the dedicated Ruby container using Ruby 4.0.6 on aarch64 Linux. A temporary stdlib benchmark used a BaseTransport whose read method returned a frozen eight-byte binary string, with 1,000,000 warmup calls followed by seven trials of 5,000,000 calls:

bundle exec ruby -Ilib /tmp/rb-read-all-benchmark.rb

• Master c2def39207a73394420088da9b4b105571dd9036: median 6,406,402 operations/second
• Proposed change: median 5,902,949 operations/second
• Difference: −7.9%

This deliberately removes all underlying I/O and therefore magnifies the cost of the EOF and progress checks. Network, file, and layered transports should spend most of their time in the underlying read. Returning a complete first chunk directly avoids the allocation and copy that an intermediate accumulator would otherwise add.

• Did you create an Apache Jira ticket? THRIFT-6144
• If a ticket exists: Does your pull request title follow the pattern "THRIFT-NNNN: describe my issue"?
• Did you squash your changes to a single commit? (not required, but preferred)
• Did you do your best to avoid breaking changes? If one was needed, did you label the Jira ticket with "Breaking-Change"?
• If your change does not involve any code, include [skip ci] anywhere in the commit message to free up build resources.

[Copilot]

Pull request overview

This PR fixes Thrift::BaseTransport#read_all in the Ruby library to properly detect end-of-input when the underlying transport returns nil or an empty string, avoiding both a leaked NoMethodError and an infinite retry loop when reads make no progress.

Changes:

• Treat nil and empty-string reads as TransportException::END_OF_FILE when additional bytes are still required.
• Preserve existing behavior for size == 0 (return an empty binary buffer without calling read) and for partial progress (continue accumulating while each read returns data).
• Add spec coverage for EOF-on-no-progress cases, plus frozen first-chunk behavior.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
lib/rb/lib/thrift/transport/base_transport.rb	Updates read_all to raise END_OF_FILE on nil/empty reads and to only accumulate while reads make progress (also ensures appended chunks are forced to binary).
lib/rb/spec/base_transport_spec.rb	Adds focused tests covering empty/nil reads, progress behavior, frozen first chunk handling, and the zero-size read fast path.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.