fix: fix compile error on macOS with clang 21

[wgtmac]

Purpose

Compiling paimon-cpp on macOS with clang 21 has some minor issues.

Tests

N/A

API and Format

N/A

Documentation

N/A

Generative AI tooling

No

[CLAassistant]

All committers have signed the CLA.

[Copilot]

Pull request overview

This PR fixes several macOS clang 21 compilation issues by resolving type-mismatch/override problems and ensuring template calls select the intended overloads in the blob/avro format code paths.

Changes:

• Fix std::min type deduction issues by explicitly selecting uint64_t and casting to uint32_t where InputStream::Read expects uint32_t.
• Fix MemoryPoolImpl::Realloc signature to correctly override MemoryPool::Realloc on platforms where size_t != uint64_t.
• Reorder Blob::FromPath definitions so Blob::Impl is complete when used with std::make_unique.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
src/paimon/format/blob/blob_format_writer.cpp	Makes read-chunk sizing compile-clean by disambiguating std::min and casting for Read(...).
src/paimon/format/avro/avro_input_stream_impl.cpp	Adjusts Read(...) sizing to match uint32_t API requirements.
src/paimon/common/memory/memory_pool.cpp	Fixes Realloc override signature for cross-platform compilation.
src/paimon/common/data/blob.cpp	Moves FromPath implementations below Blob::Impl definition to avoid incomplete-type issues.

Comments suppressed due to low confidence (1)

src/paimon/common/memory/memory_pool.cpp:62

• total_allocated_size is an atomic<int64_t>, but new_size - old_size is computed in size_t (unsigned). When shrinking, this relies on unsigned underflow and implementation-defined conversion to produce a negative delta, which is brittle and can break accounting or trigger compiler diagnostics. Compute the delta in a signed type explicitly (e.g., int64_t delta = static_cast<int64_t>(new_size) - static_cast<int64_t>(old_size)) before calling fetch_add (same applies to the similar fetch_add(new_size - old_size) in the aligned branch).

void* MemoryPoolImpl::Realloc(void* p, size_t old_size, size_t new_size, uint64_t alignment) {
    if (alignment == 0) {
        void* memptr = ::realloc(p, new_size);
        total_allocated_size.fetch_add(new_size - old_size);
        max_allocated.store(std::max(total_allocated_size.load(), max_allocated.load()));

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[lxy-9602]

+1

[wgtmac]

Thank you, @lxy-9602!

[Smith-Cruise]

@wgtmac Hi, can you compile passed on Mac?