Processing strings longer than ~2114 bytes causes a corrupted size vs. prev_size glibc malloc abort. The culprit is RuneStrArray being backed by LocalVector<struct RuneStr> — a container that explicitly documents it is only safe for primitive types, not structs.

Changes

• include/cppjieba/Unicode.hpp
  • Change RuneStrArray from limonp::LocalVector<struct RuneStr> to std::vector<RuneStr>
  • Change DecodeUTF8RunesInString loop counter from uint32_t i to size_t i to eliminate infinite-loop risk on strings >4 GB; add explicit overflow guard before narrowing casts to uint32_t fields in RuneStr

// Before
typedef limonp::LocalVector<struct RuneStr> RuneStrArray;
for (uint32_t i = 0, j = 0; i < len;) { ... }

// After
typedef std::vector<RuneStr> RuneStrArray;
for (size_t i = 0, j = 0; i < len;) {
    if (i > UINT32_MAX || j > UINT32_MAX) { runes.clear(); return false; }
    ...
}

• test/unittest/unicode_test.cpp — Update UnicodeTest.Test1 expected string to match std::vector's operator<< format (no extra element-wrapping quotes).
• test/unittest/segments_test.cpp — Add MixSegmentTest.LongInput regression test covering Chinese and ASCII inputs exceeding 2114 bytes.

• Fixes corrupted size vs. prev_size #210

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.