Cursor was set to oldest of read window, not oldest of returned
batch — each page-up jumped past entire ranges (F16: 600/2000 msgs).
Always bisect now; correct + simpler. Cost: +5-10 ms per page-up at
1M msgs, well below UI threshold.
Also guard state->entries NULL when total_lines>0 && n_entries==0
(every line failed _ff_maybe_index_line). Backup would segfault.
F16 tightened to received == total_msgs. Catches both stuck-on-
entries[0] and cursor-jumping regressions.
When the cursor lands on entries[0].byte_offset, the !from_start
backup logic produces an empty [X, X) range, get_previous_chat
returns DB_RESPONSE_EMPTY, the UI sets WIN_SCROLL_REACHED_TOP and
never tries again — page-up is permanently stuck.
Guard the cursor optimisation with two conditions: index must be
non-empty AND cursor must not equal entries[0].byte_offset. When
either fails, fall through to the end_time bisect path which
correctly locates the byte range with older messages.
F16 regression test in bench_failure_modes:
without guard: received=200 in 3 iters (premature EMPTY)
with guard: received=600 in 7 iters (truly reaches top)
Address 6 of 7 high/medium audit findings on database_export.c.
End-to-end migration of 1M rows: ~5 min -> ~1 min.
H1 cache g_slist_length(merged) out of the write loop — was O(n²)
via 2000 list walks at 500-row progress interval, now O(1).
S7a_export_cold @ 1M: 309 s -> 11.2 s (-96 %)
L1 splice existing flatfile lines into merged via ownership
transfer instead of 12-strdup deep-copy + g_date_time_ref;
free list nodes only, NULL out existing.
S7b_export_dedup @ 1M: 304 s -> 13.3 s (-96 %)
RSS @ S7b 1M: 2.4 GB -> 1.9 GB
H2 mkstemp() with random suffix instead of fixed
"{log_path}.export.tmp". Eliminates the unlink-and-retry race
where two concurrent exports could clobber each other's
in-flight tmp files.
M4 SHA-256 over full body (incremental g_checksum_update) in
the dedup-key fallback. Previous body[:256] hash collided
on common preambles (signatures, code blocks) — second
occurrence silently dropped on import.
M2 Refuse _ff_read_all_lines on files > 2 GB. Each line expands
~10x in heap on parse; without a cap we OOM-kill prof on
large flatfiles. Surface a clear error instead.
M5 Break the outer import loop on system-level INSERT failure.
Was cascading "Import of X failed" through every remaining
contact when the real issue was disk-full / sqlite locked.
Verified on bench-pipeline-max: 1M rows, full content diff,
mismatches=0. baseline.csv updated.
Deferred: M3 O_TMPFILE (platform-conditional + linkat dance;
mkstemp covers most of the same window) and M6 SIGINT cancellation
(needs event-loop architecture changes).
End-to-end performance and correctness harness for the flat-file +
SQLite database backends. Lives in tests/bench/, built only on
demand (`make bench`); not part of `make check`.
Components
gen_history (P1)
Deterministic corpus generator. Knobs: lines, contacts, years,
seed, stanza-id mode (uuid/libpurple/conversations/mixed), LMC
rate, MAM-OOO rate, resources/contact, length profile
(short/mixed/long/extreme). Emits the canonical
flatlog/<account>/<contact>/history.log layout used by
ff_verify_integrity. ~340 LOC.
bench_runner (P1, P2.5)
S1 cold tail-access via sparse index
S2 warm tail-access (page cache hot)
S3 deep pagination (1000 binary-search lookups)
S4 first-time index build (cold file -> ff_state_ensure_fresh)
S5 incremental extend (asserts no full rebuild path)
S6 real ff_verify_integrity over the contact tree
Reports total/err/warn/info issue counts in the CSV note.
bench_long_messages (P2)
L1-L14 long-message stress: 1KB up to 9.9MB bodies, plus
oversized line rejection (10MB+1 -> ff_readline returns ""),
embedded-newline / pipe / emoji body patterns, full parse on
100x1MB, 1000x100KB sustained append.
bench_failure_modes (P3)
F1-F15 failure-injection: truncated last line, mid-file CRLF,
mid-file BOM, LMC cycle, LMC depth>FF_MAX_LMC_DEPTH, manual
': ' in resource, RTL/ZWSP, Latin-1 byte, empty body,
mtime/inode flip, empty file. Each test asserts expected issue
levels and reports PASS/FAIL.
bench_export_import (P5)
Links real database_export.c + database_sqlite.c + database.c
and drives log_database_export_to_flatfile /
log_database_import_from_flatfile under load.
Subcommands: seed, export, import, roundtrip, verify.
S7a/b export, S8a/b import, S8e roundtrip with full byte-by-byte
content diff of every row in (from_jid, to_jid, message,
timestamp, type, stanza_id, archive_id, encryption, replace_id).
Make targets
bench-quick / bench / bench-full
bench-longmsg, bench-failure
bench-multicontact, bench-lmc, bench-ooo
bench-export, bench-import, bench-roundtrip
bench-pipeline, bench-pipeline-max (1M rows)
bench-compare, bench-update-baseline
Volume controls: BENCH_VOLUME (small/medium/max), BENCH_PIPE_ROWS,
BENCH_PIPE_ROWS_MAX, BENCH_DATA_DIR, BENCH_CSV.
Baseline + regression checking (P4)
tests/bench/baseline.csv 51 rows: S1-S6 x {small,lmc,ooo}
+ L1-L14 + F1-F15 (11 of 15) +
S7/S8 x {pipe100k, pipe1M}.
compare_baseline.py median over duplicate rows;
exits 1 on any (scenario, volume)
slowdown >= threshold (default 25%).
Verified at scale
bench-pipeline-max: 1,000,000 rows, full content diff
seed 17 s, export 304 s, import 31 s, idempotent re-import 10 s,
diff 2.7 s -- mismatches=0.
Findings surfaced by the harness
Export scales super-linearly: 4 s @ 100k -> 304 s @ 1M (76x for
10x rows). Cause: g_slist_sort on the merged list + per-row
ProfMessage/ff_parsed_line_t allocations. RSS peaks at 1.4 GB
on 1M. Worth a follow-up.
Export progress reporting only fires during the write phase --
the merge+sort phase (~95% of wall time at 1M) is silent.
/history export and /history import are blocking on the main
UI thread; profanity is frozen for the duration (~5 min @ 1M).
ff_readline sets *truncated=TRUE on partial-write tail but
ff_verify_integrity does not surface this -- partial writes
go unflagged (failure-injection F1).
Parser silently truncates body at first unescaped ': ' if a
resource was manually edited to contain it (F9).
In gen_history (caught by the bench's own real-verify pass):
g_strndup mid-codepoint truncation on UTF-8 bank strings -- fixed.
Linkage strategy
database_flatfile.c + parser + verify + common.c are linked
unconditionally. The export/import bench additionally links
database.c + database_sqlite.c + database_export.c. bench_stubs.c
provides minimal stubs for log_*, prefs_*, connection_get_jid,
jid_create, files_*, message_*, ui hooks. integrity_issue_free
is a weak symbol so it falls back to the real database.c
implementation when that file is linked.
Fix send_receipt_request test that wasted 60s on every run:
- Change caps hash from sha-256 to sha-1 (profanity only supports sha-1
for standard caps path, sha-256 goes to unsupported-hash fallback)
- Fix expected output: 'Buddy1 is online' → 'Buddy1 (laptop) is online'
(resource is always shown for full JID presence)
- Wrap prof_output_exact in assert_true so mismatches fail fast instead
of silently waiting the full expect_timeout
Rebalance test groups for parallel execution:
- Move Disco tests (14) from Group 3 → Group 1
- Move DB History tests (12) from Group 2 → Group 3
- Before: G1=31s G2=63s G3=30s G4=37s (bottleneck 63s)
- After: G1=46s G2=38s G3=40s G4=36s (bottleneck 46s, -27%)
Add per-test wall-clock timing via clock_gettime(CLOCK_MONOTONIC)
in init_prof_test/close_prof_test for profiling individual tests.
- Add archive_ids hash set and stanza_senders hash map to
ff_contact_state_t, populated during index build/extend via
lightweight _ff_cache_line_ids() (no full parse overhead).
- Replace O(n) file scans in _ff_has_archive_id() and
_ff_find_original_sender() with O(1) hash table lookups.
- Replace g_slist_append with g_slist_prepend + g_slist_reverse
in _ff_apply_lmc, _flatfile_get_previous_chat, and
ff_verify_integrity to avoid O(n²) list building.
- Add db_sqlite_last_changes() declaration to database.h.
Export/Import improvements:
- Replace pagination with direct SQL query (db_sqlite_get_all_chat)
- Wrap import in SQL transaction with rollback on error
- Add fsync before fclose in export for data safety
- Sort merged output by timestamp with secondary key (stanza_id, from_jid)
- Export archive_id and marked_read from SQLite (lossless migration)
- Add progress indication every 500 messages during write/import
- Expand dedup key body prefix from 64 to 256 chars
- Fix g_slist_append O(n²) → g_slist_prepend + g_slist_reverse O(n)
Field parity (to_jid, to_resource, marked_read):
- Add fields to ff_parsed_line_t struct
- Write/parse to:|to_res:|read: metadata tags in flatfile format
- Pass to_resource through _ff_add_message and all callers
- Add marked_read to ProfMessage struct with -1 default (unset)
- Preserve fields across export/import round-trips
Tests (19 new: 11 unit + 8 functional):
- Unit: to_jid_and_marked_read, bracket_in_stanza_id, backslash_in_resource,
mucpm_type, all_enc_types, crlf_handling, to_jid_special_chars,
multiple_lines, parsed_line_free_null_safe, no_space_rejected,
unclosed_bracket
- Functional: export_idempotent_no_duplicates, export_lmc_correction_survives,
switch_preserves_old_backend_data, export_all_contacts,
import_double_dedup, verify_after_export,
switch_backends_independent_messages, export_empty_contact
- Rebalance test groups: move Chat Session from Group 3 to Group 4
(25/33/30/27 instead of 25/33/36/21)
- Remove hardcoded test counts from group comments
Man page:
- Document /history switch sqlite|flatfile
Add backend-agnostic functional tests for database message persistence.
All tests work with both SQLite and flat-file backends.
New tests in test_history.c:
- message_db_history_on_reopen: basic incoming write+read round-trip
- message_db_history_multiple: 3 messages from different resources
- message_db_history_contact_isolation: buddy1 msg absent from buddy2
- message_db_history_special_chars: XML entities survive decode+DB
- message_db_history_outgoing: sent message persists across reopen
- message_db_history_dialog: outgoing + incoming both in history
- message_db_history_empty: no crash on contact with no history
- message_db_history_long_message: 1000+ char body not truncated
- message_db_history_newline: embedded LF stored correctly
- message_db_history_service_chars: backslash pipe percent braces etc
- message_db_history_verify: /history verify (ifdef HAVE_HISTORY_VERIFY)
- message_db_history_lmc: XEP-0308 correction replaces original in DB
Rebalance test groups for parallel execution (19/22/22/21):
- Group 1: Connect, Ping, Rooms, Software, LastActivity
- Group 2: Message, Receipts, Roster, DB History
- Group 3: Chat Session, Presence, Disconnect
- Group 4: MUC, Carbons
Add log_database_switch_backend() that closes the current backend,
updates PREF_DBLOG preference, and reinitializes with the new backend
without requiring a reconnect or restart.
New command: /history switch sqlite|flatfile
- Validates connection state and backend name
- Skips if already using the requested backend
- Autocomplete support for the subcommand
DRAFT — not yet tested end-to-end with a live XMPP session.
New commands:
/history export [<jid>] — copy messages from SQLite to flat-file
/history import [<jid>] — copy messages from flat-file to SQLite
Both merge with existing data; duplicates are skipped using stanza-id
or a SHA-256 fallback key (timestamp + sender + body prefix).
Implementation (src/database_export.c):
- Export: paginate SQLite via get_previous_chat, read existing flatfile
for dedup, write merged result via ff_write_line + atomic rename
- Import: parse flatfile lines via ff_parse_line, build dedup set from
SQLite, insert new messages via add_incoming (preserves original
timestamps for both directions)
- List all contacts: db_sqlite_list_contacts() queries UNION of
DISTINCT from_jid/to_jid; flatfile enumerates flatlog directories
Wiring:
- database.h: declare export/import functions + db_sqlite_list_contacts
- database_sqlite.c: add db_sqlite_list_contacts()
- cmd_defs.c: add export/import to /history synopsis and args
- cmd_funcs.c: add export/import handlers in cmd_history()
- cmd_ac.c: add 'export' and 'import' to history_ac
- Makefile.am: add database_export.c to core_sources
- stub_database.c: add stubs for test linking
- profanity.1: document export/import in man page
All code guarded with #ifdef HAVE_SQLITE — builds cleanly without it.
Build system:
- Add --without-sqlite configure flag (AM_CONDITIONAL BUILD_SQLITE)
- Guard database_sqlite.c with HAVE_SQLITE in Makefile.am, database.c,
database.h and stub_database.c
- Fall back to flatfile when SQLite not compiled in
Security (database_flatfile.c):
- MAM dedup: skip incoming messages with duplicate stanza-id (archive_id)
- LMC sender validation: reject corrections from mismatched JIDs
- Add flock() advisory locking to prevent interleaved writes from
concurrent instances
- Check fprintf return when writing file header
Autocomplete (cmd_ac.c):
- Add 'flatfile' to /privacy logging autocomplete
- Add dedicated /history autocomplete with on/off/verify (was boolean-only)
Code quality:
- Fix fwrite return type: size_t not ssize_t (database_flatfile_parser.c)
- Fix mixed allocators in ff_readline: use malloc() consistently for
overlength-line fallback instead of g_strdup()
- Fix index alignment in _ff_state_extend: use local counter so index
step doesn't depend on total_lines from initial build
Documentation:
- Fix page: correct directory structure (history.log not per-day files)
- Fix page: add aid:{archive_id} to line format example
- Fix page: missing newline before .SH BUGS
Replace per-contact directory structure with a single flat file per contact.
Each file uses pipe-delimited records with a sparse byte-offset index
that is rebuilt on demand and cached in memory.
- Rewrite _ff_write_record / _ff_read_records for single-file format
- Add sparse index (every Nth record) for O(log N) seek on history read
- Cursor-based pagination: _ff_get_previous_chat uses saved byte offset
- LMC corrections applied in-place during read (_ff_apply_lmc)
- Newline escaping (\n) in message bodies for line-based storage
- Simplify verify: parse + timestamp order + LMC reference check
- Update parser helpers for new field layout
Add pluggable storage backend abstraction (vtable) to the database layer,
allowing selection between SQLite (default) and a new flat-file backend
that stores messages as human-readable plain text files.
New files:
- database_sqlite.c: extracted SQLite backend from database.c
- database_flatfile.c: plain text backend with tolerant parser,
LMC correction chains, UTF-8/BOM/CRLF handling, integrity checks
Commands:
- /privacy logging flatfile — switch to flat-file backend
- /history verify [<jid>] — check integrity of stored history
Fixes:
- Add missing 'off' guard in flatfile backend
- Enable CHLOG/HISTORY prefs when switching to flatfile mode
Logs stored in ~/.local/share/profanity/flatlog/{account}/{contact}/{date}.log
Format: {ISO8601} [{type}|{enc}|id:{id}|corrects:{id}] {sender}: {message}
Updated: CHANGELOG, CONTRIBUTING.md, profrc.example, man page, cmd_defs,
Makefile.am, test stubs, functional test support (PROF_FLATFILE=1)
Replace 16 interactive UI setup commands with pre-written profrc file
containing [ui] and [notifications] sections (~1800ms saved per test).
Replace sleep(1) + blocking waitpid with close(pty fd) → SIGHUP →
polling waitpid(WNOHANG) → SIGTERM/SIGKILL fallback chain (~4900ms
saved per test).
Remove post-stbbr_start() and post-stbbr_stop() sleeps — bind+listen
completes synchronously before stbbr_start() returns, and
pthread_join() in stbbr_stop() guarantees socket cleanup (~200ms saved).
Add PORTS_PER_GROUP=50 isolated port ranges per test group to enable
safe parallel execution of 4 groups without port conflicts.
Add 7 new tests for /disco command:
- disco_items_to_jid: query items to specific JID
- disco_info_empty_result: handle empty disco#info response
- disco_info_multiple_identities: multiple identity elements
- disco_info_without_name: identity without optional name attr
- disco_items_without_name: items without optional name attr
- disco_info_service_unavailable: error handling for info
- disco_items_error_handling: error handling for items (XEP-0030 §7)
The disco_items_error_handling test documents a bug where disco#items
errors are silently ignored (unlike disco#info which handles them).
This violates XEP-0030 Section 7 which requires error feedback to user.
- Add 8 tests for disco info and disco items commands
- Fix XEP-0030 compliance bug: show message for empty disco#items results
- Tests cover: identity display, features, server/jid queries, error handling,
items display, empty results, and connection requirement
Per XEP-0030 section 3.1: 'if an entity has no associated items,
it MUST return an empty <query/> element.'
The client should display 'No service discovery items for X' when
receiving an empty result, not silently ignore it.
Previous version introduced in commit f28655c5c (2016-05-08) which added an early
return when child == NULL, preventing the message from being shown.
Major changes:
Run 4 build configurations in parallel with Valgrind on Linux
Add test failure detection verification (meta-test)
Port allocation per build to prevent conflicts in parallel runs
Add --coverage-only flag for dedicated coverage builds
Code quality:
Add TEST_GROUPS constant, CMOCKA patterns, helper functions
Organize ci-build.sh into sections
Remove -c http.sslverify=false from all git clones
(enables proper TLS verification, closes MITM risk).
Explicitly install ca-certificates in every CI Docker image.
Split functional tests into 4 parallel groups and add check-functional-parallel target (~3x faster CI runs).
Add branch-aware LCOV coverage reporting with new --enable-coverage option and lcov summary in CI pipeline.
Enable ccache via -C configure flag for faster recompilations.
Install lcov in all Docker images and use --depth 1 git clones + parallel make -j$(nproc) for quicker container builds.
Update CONTRIBUTING.md with instructions for parallel test groups and adding new ones.
All changes are tightly related CI/performance improvements developed in sequence. No external service uploads (e.g. Codecov skipped due to Gitea incompatibility).
- Add prof.supp with pthread TLS suppressions
- Update ci-build.sh with test configurations
- Document functional test best practices in CONTRIBUTING.md
