fix: resize Segment data_ to match capacity in ConcurrentArrayInvertedLists

[MickeyPvX]

Problem

In thirdparty/faiss/faiss/cppcontrib/knowhere/invlists/InvertedLists.h, the Segment helper inside ConcurrentArrayInvertedLists reserves capacity in its constructor but never sizes the underlying buffer:

template <typename T>
struct Segment {
    Segment(size_t segment_size, size_t code_size)
        : segment_size_(segment_size), code_size_(code_size) {
        data_.reserve(segment_size_ * code_size_);
    }
    T& operator[](idx_t idx) {
        assert(idx < segment_size_);
        return data_[idx * code_size_];
    }
    ...
    std::vector<T> data_;
};

After the constructor runs, data_.size() == 0 while data_.capacity() is segment_size_ * code_size_. The subsequent operator[] accesses (directly and via the memcpy calls in add_entries) read and write at offsets that are within the allocation but past size(). Calling std::vector::operator[] with an index that is not less than size() is undefined behavior even when the bytes are allocated.

What this breaks

On a libstdc++ built with -D_GLIBCXX_ASSERTIONS (the hardened libstdc++ shipped by several distributions), operator[] has a bounds check compiled in. The first streaming insert into an IVF_FLAT_CC or IndexIVFScalarQuantizerCC collection trips it and the process aborts:

/usr/lib/gcc/x86_64-pc-linux-gnu/12/include/c++/bits/stl_vector.h:1128:
std::vector<_Tp, _Alloc>::reference std::vector<_Tp, _Alloc>::operator[](size_type)
[with _Tp = long int; size_type = long unsigned int]: Assertion '__n < this->size()' failed.
SIGABRT: abort signal arrived during cgo execution
_ZSt21__glibcxx_assert_failPKciS0_S0_
_ZN5faiss28ConcurrentArrayInvertedLists11add_entriesEmmPKlPKhPKf
_ZN5faiss13InvertedLists9add_entryEmlPKhPKfPv
_ZN5faiss12IndexIVFFlat8add_coreElPKfS2_PKlS4_Pv

Unhardened libstdc++ compiles operator[] to plain pointer arithmetic with no check, so the same code runs without complaint.

Fix

Use resize() instead of reserve() so size() == capacity() after construction. operator[] is then in-bounds, and the memcpy writes that follow keep working unchanged (they were already assuming the storage existed).

resize() does the same allocation reserve() did, plus a zero-fill of the bytes. For the trivial T types in use here (uint8_t, idx_t, float) that's a single memset(0), dominated by the allocation cost. Segment construction is not on the hot path; inserts immediately overwrite the zero bytes via memcpy.

Backport

The same constructor exists on the v2.5.x and v2.6.x lines at the older path thirdparty/faiss/faiss/invlists/InvertedLists.h (moved to cppcontrib/knowhere/invlists/ in #1554). Same one-line change applies on both with offset adjustment.

Fixes: #1643

[sre-ci-robot]

Welcome @MickeyPvX! It looks like this is your first PR to zilliztech/knowhere 🎉

[mergify]

@MickeyPvX 🔍 Important: PR Classification Needed!

For efficient project management and a seamless review process, it's essential to classify your PR correctly. Here's how:

1. If you're fixing a bug, label it as kind/bug.
2. For small tweaks (less than 20 lines without altering any functionality), please use kind/improvement.
3. Significant changes that don't modify existing functionalities should be tagged as kind/enhancement.
4. Adjusting APIs or changing functionality? Go with kind/feature.

For any PR outside the kind/improvement category, ensure you link to the associated issue using the format: “issue: #”.

Thanks for your efforts and contribution to the community!.

[alexanderguzhva]

/kind improvement
issue: #1643

[alexanderguzhva]

/lgtm
/approve

[sre-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: alexanderguzhva, MickeyPvX

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

• OWNERS [alexanderguzhva]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment