Repository navigation
rewrite(phase1): add fragment/respqueue, convert loan to fixed-capacity, ported from c-RCP - #131
Merged
Conversation
…ty, ported from c-RCP Part of the v3.0.0 ground-up rewrite (cpp-RCP #129, ROADMAP.md Phase 17), branch rewrite/v3-from-c-rcp. Second batch of Phase 1, alongside request/sequencer/scheduler (separate PR). - fragment.hpp (NEW): cpp-RCP has never had multi-AVTPDU fragmentation -- this codebase made an explicit no-go call at ROADMAP.md milestone 52 (v2.8.0). Ported rcp::fragment::plan_count()/plan() (encode) and Reassembler (decode) from c-RCP's fragment.h/fragment.c (REQ-FRAG-*, 18 requirements). Fixed-capacity: Reassembler embeds a std::array<uint8_t, 4096> instead of c-RCP's realloc()-grown buffer -- no heap allocation at all, stronger than c-RCP's own C version. Ported c-RCP issues #614/#616's oversized-reassembly lesson as a design principle: this module stays ACF-agnostic (matching c-RCP -- the real fix there lived in mock.c's dispatch layer, not fragment.c) and always reports the true, complete reassembled length, so a future integrator (Phase 3/4 wiring, out of scope here) can check against an ACF frame ceiling before re-encoding, exactly the check c-RCP's dispatch layer had to learn to add. 30 test cases, 132 assertions. - respqueue.hpp (NEW): the per-response/ack-stream TX queue TC18 Sec12.7.9 describes -- also entirely absent before. Ported from c-RCP's respqueue.h/respqueue.c (REQ-RMAP-059/061/062/063/064/065/085): octet-budgeted FIFO, per-message max_avtpdu_size_octets ceiling, and TC18 Sec12.9.4/Sec12.9.5's evict-lowest-sequence_num overflow policy (not FIFO-oldest). kMaxEntries=64 from c-RCP's RCP_RESPQUEUE_MAX_ENTRIES, backing a fixed std::array of entry slots. 27 test cases, 405 assertions. - loan.hpp: found BufferPool::pool_ was an unbounded std::vector<unique_ptr<vector<uint8_t>>> with no capacity ceiling -- a real gap flagged in an earlier parity audit this session. Converted to a std::array<vector<uint8_t>*, 64> matching c-RCP's own RCP_LOAN_POOL_MAX_ENTRIES bound exactly (c-RCP bounds entry count only, not per-buffer size -- matched that, not a stricter policy). Once the free list is full, release now frees the buffer outright instead of growing storage further, mirroring c-RCP's pool_append() degradation. Also removed a redundant heap re-wrap the old code did on every reuse. Test count 6 -> 9 (new: pool sanity, destructor-frees- free-list, and a fixed-capacity overflow test that mutation-proves the bound). Not cleanly ported, documented in-file rather than left as a follow-up TODO (structural, not a gap to close later): c-RCP's RCP_FRAGMENT_REASM_ERR_ALLOC and its allocation-failure-injection test cases across all three modules have no C++ analog -- the fixed-capacity, no-heap design structurally eliminates the allocation-failure class, and cpp-RCP has no allocation-failure-injection seam yet (that's Phase 2's allocation-seam work). Verified: full tree rebuilds clean (221/221 targets), 55/55 ctest suites pass (100%). No endpoint file (can.hpp/spi.hpp/uart.hpp/ regmap.hpp's kOptFragmentation) touched -- wiring fragmentation into real dispatch is Phase 3/4 scope, not this pass. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Signed-off-by: Matt <47545907+SoundMatt@users.noreply.github.com>
10 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Part of the v3.0.0 ground-up rewrite (#129, ROADMAP.md Phase 17). Second batch of Phase 1, alongside
request/sequencer/scheduler(separate PR).fragment.hpp(NEW)cpp-RCP has never had multi-AVTPDU fragmentation — explicit no-go at ROADMAP.md milestone 52. Ported
plan_count()/plan()(encode) +Reassembler(decode) from c-RCP'sfragment.h/fragment.c(REQ-FRAG-*, 18 requirements). Fixed-capacity:Reassemblerembeds astd::array<uint8_t, 4096>instead of c-RCP'srealloc()-grown buffer — no heap allocation at all, stronger than c-RCP's own C version. Ported c-RCP issues #614/#616's oversized-reassembly lesson as a design principle (module stays ACF-agnostic like c-RCP, always reports the true reassembled length so a future integrator can check against an ACF frame ceiling before re-encoding). 30 test cases, 132 assertions.respqueue.hpp(NEW)TC18 §12.7.9's per-response/ack-stream TX queue — also entirely absent before. Ported from c-RCP (
REQ-RMAP-059/061/062/063/064/065/085): octet-budgeted FIFO, per-message ceiling, evict-lowest-sequence_numoverflow policy (not FIFO-oldest).kMaxEntries=64matches c-RCP's bound exactly. 27 test cases, 405 assertions.loan.hppFound
BufferPool::pool_was an unboundedstd::vector<unique_ptr<vector<uint8_t>>>— a real gap flagged in an earlier parity audit. Converted to a boundedstd::array<vector<uint8_t>*, 64>matching c-RCP'sRCP_LOAN_POOL_MAX_ENTRIESexactly. Test count 6 → 9 (incl. a fixed-capacity overflow test that mutation-proves the bound).Not cleanly ported (documented in-file, not a follow-up TODO)
c-RCP's
RCP_FRAGMENT_REASM_ERR_ALLOCand allocation-failure-injection test cases — structurally eliminated by the fixed-capacity, no-heap design; cpp-RCP has no allocation-failure-injection seam yet (Phase 2 scope).Verified
Full tree rebuilds clean (221/221 targets), 55/55 ctest suites pass (100%) — independently re-verified from a fresh build, not just the porting agent's own report. No endpoint file touched — wiring fragmentation into real dispatch is Phase 3/4 scope.
Tracked at #129.