diff options
Diffstat (limited to 'store/test-tree-store.c')
| -rw-r--r-- | store/test-tree-store.c | 1131 |
1 files changed, 1131 insertions, 0 deletions
diff --git a/store/test-tree-store.c b/store/test-tree-store.c new file mode 100644 index 0000000..1130acd --- /dev/null +++ b/store/test-tree-store.c @@ -0,0 +1,1131 @@ +#define _POSIX_C_SOURCE 200809L +#include "store.h" +#include "state.h" +#include <assert.h> +#include <dirent.h> +#include <fcntl.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <sys/stat.h> +#include <pthread.h> +#include <sys/wait.h> +#include <time.h> +#include <unistd.h> + +#define CHECK(x) do { if (!(x)) { fprintf(stderr, "%s:%d: %s\n", __FILE__, __LINE__, #x); abort(); } } while (0) + +static int fail_stage = -1; +static int crash; +static int slow_once; /* the next publication pauses before its data sync */ +static int held_sync; /* 1: the next publication waits after its head write; 2: it is waiting */ +static void pause_ms(long ms) { + struct timespec wait = {ms / 1000, (ms % 1000) * 1000000}; + nanosleep(&wait, NULL); +} +int pure_store_test_fault(PureStoreStage stage) { + if (stage == PURE_BEFORE_FILE_SYNC && __atomic_exchange_n(&slow_once, 0, __ATOMIC_SEQ_CST)) pause_ms(100); + int one = 1; + if (stage == PURE_BEFORE_DIR_SYNC && + __atomic_compare_exchange_n(&held_sync, &one, 2, false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) + while (__atomic_load_n(&held_sync, __ATOMIC_SEQ_CST) == 2) pause_ms(1); + if ((int)stage != fail_stage) return 0; + if (crash) _exit(77); + return 1; +} +bool pure_tree_test_shape(const char *path, const char *key, unsigned *height, + uint64_t *leaves, uint64_t *small, uint64_t *file_end); + +void pure_draft_test_arena(const PureDraft *draft, size_t *nodes, size_t *bytes); +unsigned long pure_tree_test_publications(void); +unsigned long pure_tree_test_syncs(void); +void pure_tree_test_forget(void); + +/* The file: head slots in blocks 0 and 1 (sequence at +8, end at +16, + * catalog at +24), commits from DATA_START, each a 128-byte record (its head) + * and then its bytes. */ +enum { KEYS = 4, CAPACITY = 1 << 20, SLOT_BLOCK = 4096, DATA_START = 8192, RECORD_SIZE = 128 }; +typedef struct { unsigned char *bytes; size_t size; int64_t revision; } Oracle; + +static uint32_t random_value(uint32_t *state) { + *state ^= *state << 13; + *state ^= *state >> 17; + *state ^= *state << 5; + return *state; +} + +static void expect(const char *path, const char *key, const Oracle *o) { + PureStoredView view; + CHECK(pure_store_get(path, key, PURE_OBJECT_LIMIT, &view) == PURE_OK); + CHECK(view.revision == o->revision && view.body.size == o->size); + CHECK(!o->size || !memcmp(view.body.data, o->bytes, o->size)); + pure_store_release(&view); +} + +static uint64_t file_size(const char *path) { + char name[512]; + snprintf(name, sizeof(name), "%s/pure.tree", path); + struct stat info; + CHECK(stat(name, &info) == 0); + return (uint64_t)info.st_size; +} + +/* Where the newest commit ends. */ +static uint64_t published_end(const char *path, const char *key) { + unsigned height; + uint64_t leaves, small, end; + CHECK(pure_tree_test_shape(path, key, &height, &leaves, &small, &end)); + return end; +} + +static uint64_t field_at(int fd, uint64_t at) { + unsigned char bytes[8]; + CHECK(pread(fd, bytes, 8, (off_t)at) == 8); + uint64_t value = 0; + for (unsigned i = 0; i < 8; ++i) value |= (uint64_t)bytes[i] << (8 * i); + return value; +} + +static uint64_t slot_field(int fd, unsigned slot, unsigned offset) { return field_at(fd, slot * SLOT_BLOCK + offset); } + +/* Where the newest head is: the newer slot's, or the last record after it. */ +static uint64_t newest_head(int fd) { + uint64_t at = slot_field(fd, 0, 8) > slot_field(fd, 1, 8) ? 0 : SLOT_BLOCK; + for (;;) { + uint64_t end = field_at(fd, at + 16); + unsigned char magic[8]; + if (pread(fd, magic, 8, (off_t)end) != 8 || memcmp(magic, "PUREREC1", 8) || + field_at(fd, end + 8) != field_at(fd, at + 8) + 1) return at; + at = end; + } +} + +static void fill(unsigned char *bytes, size_t size, uint32_t *state) { + for (size_t i = 0; i < size; ++i) bytes[i] = (unsigned char)random_value(state); +} + +/* Random edits against a flat oracle, including views held across later + * commits and compactions. Every stored node must stay balanced. */ +static void oracle(const char *path) { + static const char *keys[KEYS] = {"alpha", "b", "gamma-long_key-0123456789", "z"}; + Oracle docs[KEYS] = {{0}}; + unsigned char *insert = malloc(65536); + for (unsigned k = 0; k < KEYS; ++k) CHECK((docs[k].bytes = malloc(CAPACITY))); + CHECK(insert); + uint32_t state = 0x9e3779b9u; + PureStoredView held = {0}; + unsigned char *held_copy = malloc(CAPACITY); + size_t held_size = 0; + unsigned max_height = 0; + size_t max_size = 0, splices = 0; + for (unsigned step = 0; step < 6000; ++step) { + unsigned k = random_value(&state) % KEYS; + Oracle *o = &docs[k]; + uint32_t kind = random_value(&state) % 100; + size_t count = kind < 70 ? random_value(&state) % 24 : + kind < 95 ? random_value(&state) % 6000 : random_value(&state) % 65536; + int64_t revision; + fill(insert, count, &state); + if (kind >= 98 || !o->revision) { + CHECK(pure_store_put(path, keys[k], o->revision, (PureSpan){insert, count}, &revision) == PURE_OK); + memcpy(o->bytes, insert, count); + o->size = count; + } else { + size_t offset = o->size ? random_value(&state) % (o->size + 1) : 0; + size_t remove = o->size - offset ? random_value(&state) % (o->size - offset + 1) : 0; + if (random_value(&state) % 24) remove = remove % 64; + if (o->size - remove + count > CAPACITY) { count = 0; remove = o->size - offset; } + CHECK(pure_store_splice(path, keys[k], o->revision, offset, remove, + (PureSpan){insert, count}, PURE_OBJECT_LIMIT, &revision) == PURE_OK); + memmove(o->bytes + offset + count, o->bytes + offset + remove, o->size - offset - remove); + memcpy(o->bytes + offset, insert, count); + o->size = o->size - remove + count; + ++splices; + } + if (o->size > max_size) max_size = o->size; + CHECK(revision == o->revision + 1); + o->revision = revision; + expect(path, keys[k], o); + unsigned height; + uint64_t leaves, small, end; + CHECK(pure_tree_test_shape(path, keys[k], &height, &leaves, &small, &end)); + CHECK(small == 0); + if (height > max_height) max_height = height; + if (step % 97 == 0) { + if (held.mapped_size) { + CHECK(held.body.size == held_size && !memcmp(held.body.data, held_copy, held_size)); + pure_store_release(&held); + } + CHECK(pure_store_get(path, keys[k], PURE_OBJECT_LIMIT, &held) == PURE_OK); + held_size = held.body.size; + if (held_size) memcpy(held_copy, held.body.data, held_size); + } + } + if (held.mapped_size) { + CHECK(held.body.size == held_size && !memcmp(held.body.data, held_copy, held_size)); + pure_store_release(&held); + } + for (unsigned k = 0; k < KEYS; ++k) { expect(path, keys[k], &docs[k]); free(docs[k].bytes); } + CHECK(max_height >= 2 && max_height <= 4 && max_size > 262144); + printf("{\"oracle_steps\":6000,\"splices\":%zu,\"max_document_bytes\":%zu,\"max_height\":%u}\n", + splices, max_size, max_height); + free(insert); + free(held_copy); +} + +/* A one-byte insert, delete or replacement in a 1 MiB document appends only + * its changed leaf and path, not another copy of the document. */ +static void amplification(const char *path) { + size_t size = 1048576; + unsigned char *bytes = malloc(size + 1); + uint32_t state = 7; + CHECK(bytes); + fill(bytes, size, &state); + int64_t revision; + CHECK(pure_store_put(path, "large", 0, (PureSpan){bytes, size}, &revision) == PURE_OK); + Oracle o = {bytes, size, revision}; + unsigned char byte = 'Q'; + const size_t edits[][3] = {{size / 2, 0, 1}, {size / 3, 1, 0}, {size - 1, 1, 1}, {0, 0, 1}}; + uint64_t worst = 0; + for (size_t i = 0; i < sizeof(edits) / sizeof(*edits); ++i) { + uint64_t before = published_end(path, "large"); + size_t offset = edits[i][0], remove = edits[i][1], count = edits[i][2]; + CHECK(pure_store_splice(path, "large", o.revision, offset, remove, (PureSpan){&byte, count}, + PURE_OBJECT_LIMIT, &revision) == PURE_OK); + memmove(bytes + offset + count, bytes + offset + remove, o.size - offset - remove); + memcpy(bytes + offset, &byte, count); + o.size = o.size - remove + count; + o.revision = revision; + uint64_t after = published_end(path, "large"); + if (after > before && after - before > worst) worst = after - before; + expect(path, "large", &o); + } + /* Compaction may shrink the file; growth per edit is the claim. */ + CHECK(worst > 0 && worst < 16384); + unsigned height; + uint64_t leaves, small, end; + CHECK(pure_tree_test_shape(path, "large", &height, &leaves, &small, &end)); + CHECK(small == 0 && height == 2 && leaves >= 256 && leaves <= 512); + printf("{\"one_byte_edit_max_appended_bytes\":%llu,\"height\":%u,\"leaves\":%llu}\n", + (unsigned long long)worst, height, (unsigned long long)leaves); + free(bytes); +} + +static void faults(const char *path) { + unsigned char text[] = "fault-free"; + int64_t revision; + CHECK(pure_store_put(path, "fault", 0, (PureSpan){text, 10}, &revision) == PURE_OK); + Oracle o = {text, 10, revision}; + uint64_t size = file_size(path), end = published_end(path, "fault"); + for (fail_stage = PURE_AFTER_HEADER; fail_stage <= PURE_BEFORE_RENAME; ++fail_stage) { + CHECK(pure_store_splice(path, "fault", o.revision, 1, 3, (PureSpan){text, 5}, + PURE_OBJECT_LIMIT, &revision) == PURE_STORE_ERROR); + expect(path, "fault", &o); + /* The bytes stayed unpublished, in the written zeros after the end. */ + CHECK(published_end(path, "fault") == end && file_size(path) == size); + } + fail_stage = PURE_BEFORE_DIR_SYNC; + CHECK(pure_store_splice(path, "fault", o.revision, 10, 0, (PureSpan){text, 2}, + PURE_OBJECT_LIMIT, &revision) == PURE_UNCERTAIN); + fail_stage = -1; + unsigned char next[] = "fault-freefa"; + o = (Oracle){next, 12, o.revision + 1}; + CHECK(revision == o.revision); + expect(path, "fault", &o); +} + +/* A process that dies at each publication stage leaves either the previous + * or the next complete version; the next commit overwrites its unpublished + * bytes. A head it wrote but did not sync is made durable before the next + * writer builds on it. */ +static void crashes(const char *path) { + unsigned char text[] = "crash-base"; + int64_t revision; + CHECK(pure_store_put(path, "crash", 0, (PureSpan){text, 10}, &revision) == PURE_OK); + Oracle o = {text, 10, revision}; + for (int stage = PURE_AFTER_HEADER; stage <= PURE_BEFORE_DIR_SYNC; ++stage) { + uint64_t before = published_end(path, "crash"); + pid_t child = fork(); + CHECK(child >= 0); + if (!child) { + crash = 1; + fail_stage = stage; + pure_store_splice(path, "crash", o.revision, 0, 1, (PureSpan){(const unsigned char *)"C", 1}, + PURE_OBJECT_LIMIT, &revision); + _exit(0); + } + int status; + CHECK(waitpid(child, &status, 0) == child && WIFEXITED(status) && WEXITSTATUS(status) == 77); + if (stage == PURE_BEFORE_DIR_SYNC) { text[0] = 'C'; ++o.revision; } + expect(path, "crash", &o); + uint64_t published = published_end(path, "crash"); + CHECK(stage == PURE_BEFORE_DIR_SYNC ? published > before : published == before); + unsigned long syncs = pure_tree_test_syncs(); + CHECK(pure_store_splice(path, "crash", o.revision, 10, 0, (PureSpan){(const unsigned char *)"", 0}, + PURE_OBJECT_LIMIT, &revision) == PURE_OK); + o.revision = revision; + expect(path, "crash", &o); + CHECK(pure_tree_test_syncs() - syncs == (stage == PURE_BEFORE_DIR_SYNC ? 2u : 1u)); + CHECK(published_end(path, "crash") > published); + } +} + +static void tear(const char *path, uint64_t at) { + char name[512]; + snprintf(name, sizeof(name), "%s/pure.tree", path); + int fd = open(name, O_RDWR); + unsigned char torn; + CHECK(fd >= 0 && pread(fd, &torn, 1, (off_t)at) == 1); + torn ^= 0x5a; + CHECK(pwrite(fd, &torn, 1, (off_t)at) == 1 && close(fd) == 0); +} + +/* A torn newest record exposes the previous commit to a new process, and the + * next writer overwrites that record and the commit's bytes. A torn slot + * costs only a longer walk: the other slot's head leads along the records. */ +static void torn_head(const char *path) { + unsigned char first[] = "first", second[] = "second"; + int64_t revision; + CHECK(pure_store_put(path, "torn", 0, (PureSpan){first, 5}, &revision) == PURE_OK); + Oracle o = {first, 5, revision}; + uint64_t record = published_end(path, "torn"); /* where the next commit's record goes */ + CHECK(pure_store_put(path, "torn", o.revision, (PureSpan){second, 6}, &revision) == PURE_OK); + tear(path, SLOT_BLOCK + 20); + pure_tree_test_forget(); + expect(path, "torn", &(Oracle){second, 6, revision}); + tear(path, record + 20); + pure_tree_test_forget(); + expect(path, "torn", &o); + CHECK(pure_store_put(path, "torn", o.revision + 1, (PureSpan){first, 1}, &revision) == PURE_CONFLICT); + CHECK(pure_store_put(path, "torn", o.revision, (PureSpan){second, 3}, &revision) == PURE_OK); + o = (Oracle){second, 3, revision}; + CHECK(published_end(path, "torn") < record + RECORD_SIZE + 4096); /* it went where the torn one was */ + pure_tree_test_forget(); + expect(path, "torn", &o); +} + +/* Commits are found along their records: from a slot's head, which every + * CHECKPOINT commits moves on, by a new process too; a damaged record ends + * the walk at the commit before it, which the next writer builds on. */ +static void records(const char *path) { + int64_t revision; + uint64_t ends[41]; + char text[16]; + CHECK(pure_store_put(path, "walk", 0, (PureSpan){(const unsigned char *)"0", 1}, &revision) == PURE_OK); + for (unsigned i = 1; i <= 40; ++i) { + ends[i] = published_end(path, "walk"); + snprintf(text, sizeof(text), "%u", i); + CHECK(pure_store_put(path, "walk", revision, (PureSpan){(const unsigned char *)text, strlen(text)}, + &revision) == PURE_OK); + } + char name[512]; + snprintf(name, sizeof(name), "%s/pure.tree", path); + int fd = open(name, O_RDONLY); + CHECK(fd >= 0); + uint64_t newest_slot = slot_field(fd, 0, 8) > slot_field(fd, 1, 8) ? slot_field(fd, 0, 8) : slot_field(fd, 1, 8); + CHECK(close(fd) == 0); + /* 42 commits (the empty file's first): heads in slots at 17 and 33. */ + CHECK(newest_slot == 33); + pure_tree_test_forget(); + expect(path, "walk", &(Oracle){(unsigned char *)"40", 2, 41}); + pid_t child = fork(); + CHECK(child >= 0); + if (!child) { + pure_tree_test_forget(); /* a process that never saw them */ + PureStoredView view; + _exit(pure_store_get(path, "walk", 16, &view) == PURE_OK && view.body.size == 2 && + !memcmp(view.body.data, "40", 2) && view.revision == 41 ? 0 : 1); + } + int status; + CHECK(waitpid(child, &status, 0) == child && WIFEXITED(status) && WEXITSTATUS(status) == 0); + /* Commit 37 ("37", revision 38) has a damaged record: the walk ends at 36. */ + tear(path, ends[37] + 30); + pure_tree_test_forget(); + expect(path, "walk", &(Oracle){(unsigned char *)"36", 2, 37}); + CHECK(pure_store_put(path, "walk", 37, (PureSpan){(const unsigned char *)"x", 1}, &revision) == PURE_OK); + CHECK(revision == 38 && published_end(path, "walk") < ends[38]); + pure_tree_test_forget(); + expect(path, "walk", &(Oracle){(unsigned char *)"x", 1, 38}); +} + +/* A reader's view stays byte-identical while another process commits and + * compacts; readers take no lock. */ +static void isolation(const char *path) { + size_t size = 300000; + unsigned char *bytes = malloc(size); + uint32_t state = 99; + CHECK(bytes); + fill(bytes, size, &state); + int64_t revision; + CHECK(pure_store_put(path, "shared", 0, (PureSpan){bytes, size}, &revision) == PURE_OK); + PureStoredView view; + CHECK(pure_store_get(path, "shared", size, &view) == PURE_OK); + uint64_t before = file_size(path); + pid_t child = fork(); + CHECK(child >= 0); + if (!child) { + int64_t at = revision; + unsigned char byte = 'x'; + for (unsigned i = 0; i < 400; ++i) + if (pure_store_splice(path, "shared", at, (i * 7919u) % size, 1, (PureSpan){&byte, 1}, + PURE_OBJECT_LIMIT, &at) != PURE_OK) _exit(1); + _exit(0); + } + int status; + CHECK(waitpid(child, &status, 0) == child && WIFEXITED(status) && WEXITSTATUS(status) == 0); + CHECK(view.revision == revision && view.body.size == size && !memcmp(view.body.data, bytes, size)); + pure_store_release(&view); + PureStoredView latest; + CHECK(pure_store_get(path, "shared", size, &latest) == PURE_OK && latest.revision == revision + 400); + pure_store_release(&latest); + /* About 2 MB of path copies without compaction would exceed this. */ + CHECK(file_size(path) < before + 3 * 524288); + free(bytes); +} + +/* Damaged branch bytes are reported as storage errors, never trusted. */ +static void corruption(const char *path) { + unsigned char *bytes = malloc(200000); + uint32_t state = 3; + CHECK(bytes); + fill(bytes, 200000, &state); + int64_t revision; + CHECK(pure_store_put(path, "damaged", 0, (PureSpan){bytes, 200000}, &revision) == PURE_OK); + char name[512]; + snprintf(name, sizeof(name), "%s/pure.tree", path); + uint64_t end = published_end(path, "damaged"); + int fd = open(name, O_RDWR); + CHECK(fd >= 0); + unsigned errors = 0; + for (unsigned i = 0; i < 3000; ++i) { + off_t at = (off_t)(DATA_START + random_value(&state) % (end - DATA_START)); + unsigned char original, damaged; + CHECK(pread(fd, &original, 1, at) == 1); + damaged = original ^ (unsigned char)(1u << (random_value(&state) % 8)); + CHECK(pwrite(fd, &damaged, 1, at) == 1); + PureStoredView view; + PureResult result = pure_store_get(path, "damaged", PURE_OBJECT_LIMIT, &view); + CHECK(result == PURE_OK || result == PURE_STORE_ERROR || result == PURE_NOT_FOUND); + if (result == PURE_OK) pure_store_release(&view); else ++errors; + CHECK(pwrite(fd, &original, 1, at) == 1); + } + CHECK(close(fd) == 0 && errors > 0); + free(bytes); +} + +static void many_keys(const char *path) { + int64_t revision; + char key[32]; + for (unsigned i = 0; i < 1024; ++i) { + unsigned id = (i * 617u) % 1024u; + snprintf(key, sizeof(key), "doc-%04u", id); + CHECK(pure_store_put(path, key, 0, (PureSpan){(const unsigned char *)key, strlen(key)}, &revision) == PURE_OK); + } + for (unsigned id = 0; id < 1024; ++id) { + snprintf(key, sizeof(key), "doc-%04u", id); + Oracle o = {(unsigned char *)key, strlen(key), 1}; + expect(path, key, &o); + } + PureStoredView view; + CHECK(pure_store_get(path, "doc-1024", 64, &view) == PURE_NOT_FOUND); + CHECK(pure_store_get(path, "doc-0001", 3, &view) == PURE_BUDGET); + CHECK(pure_store_splice(path, "doc-0002", 1, 9, 0, (PureSpan){0}, PURE_OBJECT_LIMIT, &revision) == PURE_LIMIT); + CHECK(pure_store_put(path, "doc-0003", 0, (PureSpan){0}, &revision) == PURE_CONFLICT && revision == 1); +} + +/* The newest catalog of a one-key store is one leaf; forge its revision. */ +static void revision_limit(const char *path) { + int64_t revision; + CHECK(pure_store_put(path, "work", 0, (PureSpan){(const unsigned char *)"w", 1}, &revision) == PURE_OK); + char name[512]; + snprintf(name, sizeof(name), "%s/pure.tree", path); + int fd = open(name, O_RDWR); + CHECK(fd >= 0); + uint64_t catalog = field_at(fd, newest_head(fd) + 24); + CHECK(catalog >> 63); + const unsigned char maximum[] = {255,255,255,255,255,255,255,127}; + off_t record = (off_t)(catalog & ((UINT64_C(1) << 62) - 1)); + CHECK(pwrite(fd, maximum, 8, record + 80) == 8 && close(fd) == 0); + /* The forged commit no longer matches its digest; this process wrote it + * and does not read it again, a new process would pass over it. */ + PureStoredView view; + CHECK(pure_store_get(path, "work", 1, &view) == PURE_OK && view.revision == INT64_MAX); + pure_store_release(&view); + CHECK(pure_store_splice(path, "work", INT64_MAX, 0, 0, (PureSpan){0}, PURE_OBJECT_LIMIT, &revision) == PURE_LIMIT); +} + +static bool count_key(void *context, const char *key) { + (void)key; + ++*(unsigned *)context; + return true; +} + +static void keys(const char *path) { + unsigned count = 0; + CHECK(pure_store_keys(path, count_key, &count) == PURE_OK && count == 0); + int64_t revision; + const char *names[] = {"m", "a", "zz", "a-b", "A"}; + for (unsigned i = 0; i < 5; ++i) + CHECK(pure_store_put(path, names[i], 0, (PureSpan){0}, &revision) == PURE_OK); + CHECK(pure_store_keys(path, count_key, &count) == PURE_OK && count == 5); +} + +typedef struct { unsigned char *out; size_t spans, largest, limit; } Gather; + +static bool gather(void *context, PureSpan span) { + Gather *g = context; + memcpy(g->out, span.data, span.size); + g->out += span.size; + if (span.size > g->largest) g->largest = span.size; + return ++g->spans != g->limit; +} + +static void expect_draft(const PureDraft *draft, const Oracle *o, unsigned char *scratch) { + Gather g = {scratch, 0, 0, 0}; + CHECK(pure_draft_size(draft) == o->size && pure_draft_revision(draft) == o->revision); + CHECK(pure_draft_read(draft, 0, o->size, gather, &g) == PURE_OK); + CHECK((size_t)(g.out - scratch) == o->size && (!o->size || !memcmp(scratch, o->bytes, o->size))); + CHECK(g.largest <= 4096); +} + +static void apply(Oracle *o, size_t offset, size_t remove, const unsigned char *insert, size_t count) { + memmove(o->bytes + offset + count, o->bytes + offset + remove, o->size - offset - remove); + if (count) memcpy(o->bytes + offset, insert, count); + o->size = o->size - remove + count; +} + +/* Edits live only in the draft until commit; each commit continues from the + * published version with an empty arena. */ +static void drafts(const char *path) { + Oracle o = {malloc(CAPACITY), 0, 0}; + unsigned char *scratch = malloc(CAPACITY), *insert = malloc(65536); + CHECK(o.bytes && scratch && insert); + PureDraft *draft; + PureStoredView view; + CHECK(pure_draft_open(path, "draft", &draft) == PURE_OK); + expect_draft(draft, &o, scratch); + uint32_t state = 0x1234567u; + for (unsigned step = 0; step < 20000; ++step) { + uint32_t kind = random_value(&state) % 100; + size_t count = kind < 80 ? random_value(&state) % 8 : + kind < 98 ? random_value(&state) % 4000 : random_value(&state) % 65536; + size_t offset = o.size ? random_value(&state) % (o.size + 1) : 0; + size_t remove = o.size - offset ? random_value(&state) % (o.size - offset + 1) : 0; + if (random_value(&state) % 64) remove %= 16; + if (o.size - remove + count > CAPACITY) count = 0; + fill(insert, count, &state); + CHECK(pure_draft_splice(draft, offset, remove, (PureSpan){insert, count}) == PURE_OK); + apply(&o, offset, remove, insert, count); + if (step % 97 == 0) expect_draft(draft, &o, scratch); + if (step % 1000 == 999) { + if (o.revision) CHECK(pure_store_get(path, "draft", PURE_OBJECT_LIMIT, &view) == PURE_OK && + view.revision == o.revision); + if (o.revision) pure_store_release(&view); + int64_t revision; + CHECK(pure_draft_commit(draft, &revision) == PURE_OK && revision == o.revision + 1); + o.revision = revision; + expect(path, "draft", &o); + expect_draft(draft, &o, scratch); + size_t nodes, bytes; + pure_draft_test_arena(draft, &nodes, &bytes); + CHECK(nodes == 0); + } + } + CHECK(o.size > 100000); + Gather g = {scratch, 0, 0, 3}; + CHECK(pure_draft_read(draft, 0, o.size, gather, &g) == PURE_OK && g.spans == 3); + CHECK(pure_draft_splice(draft, o.size + 1, 0, (PureSpan){0}) == PURE_LIMIT); + CHECK(pure_draft_splice(draft, 0, o.size + 1, (PureSpan){0}) == PURE_LIMIT); + CHECK(pure_draft_read(draft, o.size, 1, gather, &g) == PURE_LIMIT); + expect_draft(draft, &o, scratch); + pure_draft_close(draft); + printf("{\"draft_steps\":20000,\"draft_commits\":%lld,\"draft_bytes\":%zu}\n", + (long long)o.revision, o.size); + free(o.bytes); + free(scratch); + free(insert); +} + +/* Long typing and deleting at one place keeps only the live path in memory, + * and its commit appends the typed bytes and path, not the document. */ +static void typing(const char *path) { + size_t size = 1048576; + Oracle o = {malloc(size + 40000), size, 0}; + unsigned char *scratch = malloc(size + 40000); + uint32_t state = 5; + CHECK(o.bytes && scratch); + fill(o.bytes, size, &state); + CHECK(pure_store_put(path, "typed", 0, (PureSpan){o.bytes, size}, &o.revision) == PURE_OK); + PureDraft *draft; + CHECK(pure_draft_open(path, "typed", &draft) == PURE_OK); + size_t cursor = size / 2, nodes, bytes, peak = 0; + for (unsigned i = 0; i < 30000; ++i) { + unsigned char byte = (unsigned char)('a' + i % 26); + if (i % 3 == 2) { + CHECK(pure_draft_splice(draft, cursor - 1, 1, (PureSpan){0}) == PURE_OK); + apply(&o, --cursor, 1, NULL, 0); + } else { + CHECK(pure_draft_splice(draft, cursor, 0, (PureSpan){&byte, 1}) == PURE_OK); + apply(&o, cursor++, 0, &byte, 1); + } + pure_draft_test_arena(draft, &nodes, &bytes); + if (bytes > peak) peak = bytes; + } + expect_draft(draft, &o, scratch); + /* Reachable: 10,000 net typed bytes, the leaves they landed in and one + * path. Unreachable copies are reclaimed past max(reachable, 32 KiB). */ + CHECK(peak < 2 * (10000 + 4 * 4096 + 2048) + 32768 + 8192); + uint64_t before = published_end(path, "typed"); + int64_t revision; + CHECK(pure_draft_commit(draft, &revision) == PURE_OK); + o.revision = revision; + uint64_t appended = published_end(path, "typed") - before; + CHECK(appended < 10000 + 16384); + expect(path, "typed", &o); + pure_draft_close(draft); + printf("{\"typing_keystrokes\":30000,\"peak_arena_bytes\":%zu,\"commit_appended_bytes\":%llu}\n", + peak, (unsigned long long)appended); + free(o.bytes); + free(scratch); +} + +/* Two drafts of one key: the second commit conflicts and keeps its edits. */ +static void draft_conflict(const char *path) { + unsigned char scratch[64]; + int64_t revision; + CHECK(pure_store_put(path, "shared", 0, (PureSpan){(const unsigned char *)"base", 4}, &revision) == PURE_OK); + PureDraft *first, *second; + CHECK(pure_draft_open(path, "shared", &first) == PURE_OK && pure_draft_open(path, "shared", &second) == PURE_OK); + CHECK(pure_draft_splice(first, 4, 0, (PureSpan){(const unsigned char *)"-one", 4}) == PURE_OK); + CHECK(pure_draft_splice(second, 0, 0, (PureSpan){(const unsigned char *)"two-", 4}) == PURE_OK); + CHECK(pure_draft_commit(first, &revision) == PURE_OK && revision == 2); + CHECK(pure_draft_commit(second, &revision) == PURE_CONFLICT && revision == 2); + Oracle kept = {(unsigned char *)"two-base", 8, 1}; + expect_draft(second, &kept, scratch); + Oracle stored = {(unsigned char *)"base-one", 8, 2}; + expect(path, "shared", &stored); + pure_draft_close(second); + CHECK(pure_draft_open(path, "shared", &second) == PURE_OK); + expect_draft(second, &stored, scratch); + CHECK(pure_store_put(path, "shared", 2, (PureSpan){(const unsigned char *)"x", 1}, &revision) == PURE_OK); + CHECK(pure_draft_commit(first, &revision) == PURE_CONFLICT && revision == 3); + pure_draft_close(first); + pure_draft_close(second); +} + +/* Another key's writes compact the file under an open draft; the draft's + * stored extents are copied from its old mapping on commit. */ +static void draft_relocation(const char *path) { + unsigned char *big = malloc(300000), scratch[20000]; + uint32_t state = 11; + CHECK(big); + fill(big, 300000, &state); + Oracle o = {malloc(20000), 12000, 0}; + CHECK(o.bytes); + fill(o.bytes, o.size, &state); + CHECK(pure_store_put(path, "kept", 0, (PureSpan){o.bytes, o.size}, &o.revision) == PURE_OK); + PureDraft *draft; + CHECK(pure_draft_open(path, "kept", &draft) == PURE_OK); + CHECK(pure_draft_splice(draft, 6000, 10, (PureSpan){(const unsigned char *)"relocated", 9}) == PURE_OK); + apply(&o, 6000, 10, (const unsigned char *)"relocated", 9); + char name[512]; + snprintf(name, sizeof(name), "%s/pure.tree", path); + struct stat before, after; + CHECK(stat(name, &before) == 0); + int64_t revision = 0; + for (unsigned i = 0; i < 8; ++i) + CHECK(pure_store_put(path, "other", revision, (PureSpan){big, 300000 - i}, &revision) == PURE_OK); + CHECK(stat(name, &after) == 0 && after.st_ino != before.st_ino); + CHECK(pure_draft_commit(draft, &revision) == PURE_OK && revision == o.revision + 1); + o.revision = revision; + expect(path, "kept", &o); + expect_draft(draft, &o, scratch); + Oracle other = {big, 300000 - 7, 8}; + expect(path, "other", &other); + pure_draft_close(draft); + free(o.bytes); + free(big); +} + +/* A held draft keeps earlier versions readable across later edits. */ +static void draft_versions(const char *path) { + static unsigned char bytes[200000], scratch[200000]; + uint32_t state = 17; + fill(bytes, sizeof(bytes), &state); + int64_t revision; + CHECK(pure_store_put(path, "versioned", 0, (PureSpan){bytes, sizeof(bytes)}, &revision) == PURE_OK); + PureDraft *draft; + CHECK(pure_draft_open(path, "versioned", &draft) == PURE_OK); + pure_draft_hold(draft, true); + PureVersion first = pure_draft_version(draft); + for (unsigned i = 0; i < 400; ++i) { + unsigned char byte = (unsigned char)i; + CHECK(pure_draft_splice(draft, (i * 977u) % 150000, 3, (PureSpan){&byte, 1}) == PURE_OK); + } + Gather g = {scratch, 0, 0, 0}; + CHECK(pure_draft_read_version(draft, first, 0, sizeof(bytes), gather, &g) == PURE_OK); + CHECK((size_t)(g.out - scratch) == sizeof(bytes) && !memcmp(scratch, bytes, sizeof(bytes))); + CHECK(pure_draft_size(draft) == sizeof(bytes) - 800); + size_t nodes, held_bytes, after; + pure_draft_test_arena(draft, &nodes, &held_bytes); + pure_draft_hold(draft, false); + pure_draft_test_arena(draft, &nodes, &after); + CHECK(after < held_bytes); + pure_draft_close(draft); +} + +/* A draft without a store: edits, versions, restore and reclamation. */ +static void anonymous_draft(const char *path) { + (void)path; + Oracle o = {malloc(CAPACITY), 0, 0}; + unsigned char *scratch = malloc(CAPACITY), insert[3000]; + PureDraft *draft = pure_draft_new(); + CHECK(draft && o.bytes && scratch && pure_draft_size(draft) == 0 && pure_draft_revision(draft) == 0); + uint32_t state = 42; + for (unsigned step = 0; step < 4000; ++step) { + size_t count = random_value(&state) % (step % 50 ? 12 : sizeof(insert)); + size_t offset = o.size ? random_value(&state) % (o.size + 1) : 0; + size_t remove = o.size - offset ? random_value(&state) % (o.size - offset + 1) % 32 : 0; + fill(insert, count, &state); + bool abandon = random_value(&state) % 5 == 0; + pure_draft_hold(draft, true); + PureVersion before = pure_draft_version(draft); + CHECK(pure_draft_splice(draft, offset, remove, (PureSpan){insert, count}) == PURE_OK); + if (abandon) pure_draft_restore(draft, before); + else apply(&o, offset, remove, insert, count); + pure_draft_hold(draft, false); + if (step % 37 == 0) expect_draft(draft, &o, scratch); + } + expect_draft(draft, &o, scratch); + int64_t revision; + CHECK(pure_draft_commit(draft, &revision) == PURE_INVALID); + size_t nodes, bytes; + pure_draft_test_arena(draft, &nodes, &bytes); + CHECK(bytes < 2 * o.size + 65536); + pure_draft_close(draft); + free(o.bytes); + free(scratch); +} + +/* A failed commit changes neither the store nor the draft. */ +static void draft_faults(const char *path) { + unsigned char scratch[64]; + int64_t revision; + CHECK(pure_store_put(path, "faulty", 0, (PureSpan){(const unsigned char *)"stable", 6}, &revision) == PURE_OK); + PureDraft *draft; + CHECK(pure_draft_open(path, "faulty", &draft) == PURE_OK); + CHECK(pure_draft_splice(draft, 6, 0, (PureSpan){(const unsigned char *)"+edit", 5}) == PURE_OK); + Oracle stored = {(unsigned char *)"stable", 6, 1}, edited = {(unsigned char *)"stable+edit", 11, 1}; + uint64_t size = file_size(path); + for (fail_stage = PURE_AFTER_HEADER; fail_stage <= PURE_BEFORE_RENAME; ++fail_stage) { + CHECK(pure_draft_commit(draft, &revision) == PURE_STORE_ERROR); + expect(path, "faulty", &stored); + expect_draft(draft, &edited, scratch); + CHECK(file_size(path) == size); + } + fail_stage = PURE_BEFORE_DIR_SYNC; + CHECK(pure_draft_commit(draft, &revision) == PURE_UNCERTAIN && revision == 2); + fail_stage = -1; + expect_draft(draft, &edited, scratch); + CHECK(pure_draft_commit(draft, &revision) == PURE_CONFLICT && revision == 2); + pure_draft_close(draft); + CHECK(pure_draft_open(path, "faulty", &draft) == PURE_OK); + edited.revision = 2; + expect_draft(draft, &edited, scratch); + CHECK(pure_draft_splice(draft, 0, 6, (PureSpan){0}) == PURE_OK); + CHECK(pure_draft_commit(draft, &revision) == PURE_OK && revision == 3); + Oracle final = {(unsigned char *)"+edit", 5, 3}; + expect(path, "faulty", &final); + pure_draft_close(draft); + CHECK(pure_draft_open(path, "bad/key", &draft) == PURE_INVALID && !draft); +} + +static bool list_key(void *context, const char *key) { + char *seen = context; + strcat(seen, key); + strcat(seen, ","); + return true; +} + +/* Several drafts publish in one commit: all or none, under process crashes, + * faults and a stale draft. Runtime values stay out of the document keys. */ +static void batch(const char *path) { + unsigned char scratch[64]; + int64_t revision, revisions[3]; + CHECK(pure_store_put(path, "alpha", 0, (PureSpan){(const unsigned char *)"a0", 2}, &revision) == PURE_OK); + CHECK(pure_store_put(path, "beta", 0, (PureSpan){(const unsigned char *)"b0", 2}, &revision) == PURE_OK); + PureDraft *drafts[3]; + CHECK(pure_draft_open(path, "alpha", &drafts[0]) == PURE_OK && pure_draft_open(path, "beta", &drafts[1]) == PURE_OK); + CHECK(pure_state_open(path, "values", &drafts[2]) == PURE_OK && pure_draft_revision(drafts[2]) == 0); + PureDraft *bad; + CHECK(pure_state_open(path, "bad.name", &bad) == PURE_INVALID && !bad); + CHECK(pure_draft_splice(drafts[0], 2, 0, (PureSpan){(const unsigned char *)"+1", 2}) == PURE_OK); + CHECK(pure_draft_splice(drafts[1], 0, 2, (PureSpan){(const unsigned char *)"B", 1}) == PURE_OK); + CHECK(pure_draft_splice(drafts[2], 0, 0, (PureSpan){(const unsigned char *)"state", 5}) == PURE_OK); + Oracle alpha = {(unsigned char *)"a0", 2, 1}, beta = {(unsigned char *)"b0", 2, 1}; + Oracle alpha_new = {(unsigned char *)"a0+1", 4, 1}, beta_new = {(unsigned char *)"B", 1, 1}; + Oracle state_new = {(unsigned char *)"state", 5, 0}; + /* A process that dies anywhere in the commit leaves all three as before, + * or all three published once the head is written. */ + for (int stage = PURE_AFTER_HEADER; stage <= PURE_BEFORE_DIR_SYNC; ++stage) { + pid_t child = fork(); + CHECK(child >= 0); + if (!child) { + crash = 1; + fail_stage = stage; + pure_draft_commit_all(drafts, 3, revisions); + _exit(0); + } + int status; + CHECK(waitpid(child, &status, 0) == child && WIFEXITED(status) && WEXITSTATUS(status) == 77); + PureDraft *state; + CHECK(pure_state_open(path, "values", &state) == PURE_OK); + if (stage < PURE_BEFORE_DIR_SYNC) { + expect(path, "alpha", &alpha); + expect(path, "beta", &beta); + CHECK(pure_draft_revision(state) == 0 && pure_draft_size(state) == 0); + pure_draft_close(state); + continue; + } + Oracle a = alpha_new, b = beta_new, v = state_new; + a.revision = b.revision = 2; + v.revision = 1; + expect(path, "alpha", &a); + expect(path, "beta", &b); + expect_draft(state, &v, scratch); + pure_draft_close(state); + /* Undo the crashed commit's effect for the next checks. */ + CHECK(pure_store_put(path, "alpha", 2, (PureSpan){(const unsigned char *)"a0", 2}, &revision) == PURE_OK); + } + /* The same batch in this process is now stale on every key. */ + CHECK(pure_draft_commit_all(drafts, 3, revisions) == PURE_CONFLICT && revisions[0] == 3); + for (unsigned i = 0; i < 3; ++i) pure_draft_close(drafts[i]); + CHECK(pure_draft_open(path, "alpha", &drafts[0]) == PURE_OK && pure_draft_open(path, "beta", &drafts[1]) == PURE_OK); + CHECK(pure_state_open(path, "values", &drafts[2]) == PURE_OK); + CHECK(pure_draft_splice(drafts[0], 0, 2, (PureSpan){(const unsigned char *)"A", 1}) == PURE_OK); + CHECK(pure_draft_splice(drafts[1], 1, 0, (PureSpan){(const unsigned char *)"!", 1}) == PURE_OK); + CHECK(pure_draft_splice(drafts[2], 5, 0, (PureSpan){(const unsigned char *)"+", 1}) == PURE_OK); + alpha = (Oracle){(unsigned char *)"a0", 2, 3}; + beta = (Oracle){(unsigned char *)"B", 1, 2}; + uint64_t size = file_size(path); + for (fail_stage = PURE_AFTER_HEADER; fail_stage <= PURE_BEFORE_RENAME; ++fail_stage) { + CHECK(pure_draft_commit_all(drafts, 3, revisions) == PURE_STORE_ERROR); + expect(path, "alpha", &alpha); + expect(path, "beta", &beta); + CHECK(file_size(path) == size); + } + fail_stage = -1; + /* One stale draft stops the whole batch. */ + PureDraft *other; + CHECK(pure_draft_open(path, "beta", &other) == PURE_OK); + CHECK(pure_draft_splice(other, 0, 0, (PureSpan){(const unsigned char *)"x", 1}) == PURE_OK); + CHECK(pure_draft_commit(other, &revision) == PURE_OK && revision == 3); + pure_draft_close(other); + CHECK(pure_draft_commit_all(drafts, 3, revisions) == PURE_CONFLICT && revisions[1] == 3); + expect(path, "alpha", &alpha); + pure_draft_close(drafts[1]); + CHECK(pure_draft_open(path, "beta", &drafts[1]) == PURE_OK); + CHECK(pure_draft_splice(drafts[1], 0, 0, (PureSpan){(const unsigned char *)"y", 1}) == PURE_OK); + CHECK(pure_draft_commit_all(drafts, 3, revisions) == PURE_OK); + CHECK(revisions[0] == 4 && revisions[1] == 4 && revisions[2] == 2); + expect(path, "alpha", &(Oracle){(unsigned char *)"A", 1, 4}); + expect(path, "beta", &(Oracle){(unsigned char *)"yxB", 3, 4}); + expect_draft(drafts[2], &(Oracle){(unsigned char *)"state+", 6, 2}, scratch); + /* Drafts continue from the new revisions. */ + CHECK(pure_draft_splice(drafts[0], 1, 0, (PureSpan){(const unsigned char *)"z", 1}) == PURE_OK); + CHECK(pure_draft_commit_all(drafts, 1, revisions) == PURE_OK && revisions[0] == 5); + /* Invalid batches change nothing. */ + PureDraft *twice[2] = {drafts[0], drafts[0]}, *anonymous = pure_draft_new(); + CHECK(pure_draft_commit_all(twice, 2, revisions) == PURE_INVALID); + PureDraft *mixed[2] = {drafts[1], anonymous}; + CHECK(pure_draft_commit_all(mixed, 2, revisions) == PURE_INVALID); + pure_draft_close(anonymous); + for (unsigned i = 0; i < 3; ++i) pure_draft_close(drafts[i]); + /* Runtime values are not documents: listings skip them and document + * operations refuse their names. */ + char seen[256] = ""; + CHECK(pure_store_keys(path, list_key, seen) == PURE_OK && !strcmp(seen, "alpha,beta,")); + CHECK(pure_store_put(path, ".values", 0, (PureSpan){(const unsigned char *)"x", 1}, &revision) != PURE_OK); + /* Many keys, in no particular order, in one commit. */ + PureDraft *many[50]; + int64_t many_revisions[50]; + char key[16]; + for (unsigned i = 0; i < 50; ++i) { + snprintf(key, sizeof(key), "k%02u", (i * 37) % 50); + CHECK((i % 2 ? pure_state_open(path, key, &many[i]) : pure_draft_open(path, key, &many[i])) == PURE_OK); + CHECK(pure_draft_splice(many[i], 0, 0, (PureSpan){(const unsigned char *)key, strlen(key)}) == PURE_OK); + } + CHECK(pure_draft_commit_all(many, 50, many_revisions) == PURE_OK); + for (unsigned i = 0; i < 50; ++i) { + CHECK(many_revisions[i] == 1); + snprintf(key, sizeof(key), "k%02u", (i * 37) % 50); + PureDraft *check; + CHECK((i % 2 ? pure_state_open(path, key, &check) : pure_draft_open(path, key, &check)) == PURE_OK); + expect_draft(check, &(Oracle){(unsigned char *)key, strlen(key), 1}, scratch); + pure_draft_close(check); + pure_draft_close(many[i]); + } +} + +/* A store handle shares one mapping among its drafts, sees commits made by + * other writers when a draft opens, follows a compaction that replaced the + * file, and maps a batch it committed once. */ +static void store_handle(const char *path) { + unsigned char scratch[64]; + int64_t revision, revisions[3]; + CHECK(pure_store_put(path, "shared", 0, (PureSpan){(const unsigned char *)"one", 3}, &revision) == PURE_OK); + PureStore *store; + CHECK(pure_store_open(path, &store) == PURE_OK); + PureDraft *a, *b, *state, *bad; + CHECK(pure_store_draft(store, "shared", &a) == PURE_OK); + expect_draft(a, &(Oracle){(unsigned char *)"one", 3, 1}, scratch); + CHECK(pure_store_draft(store, "bad.key", &bad) == PURE_INVALID && !bad); + /* Another writer commits: the next draft sees it, the open one keeps its version. */ + CHECK(pure_store_put(path, "shared", 1, (PureSpan){(const unsigned char *)"two", 3}, &revision) == PURE_OK); + CHECK(pure_store_draft(store, "shared", &b) == PURE_OK); + expect_draft(b, &(Oracle){(unsigned char *)"two", 3, 2}, scratch); + expect_draft(a, &(Oracle){(unsigned char *)"one", 3, 1}, scratch); + CHECK(pure_draft_splice(a, 3, 0, (PureSpan){(const unsigned char *)"!", 1}) == PURE_OK); + CHECK(pure_draft_commit(a, &revision) == PURE_CONFLICT && revision == 2); + pure_draft_close(a); + /* A batch of handle drafts, a runtime value among them. */ + CHECK(pure_store_state(store, "values", &state) == PURE_OK && pure_draft_revision(state) == 0); + CHECK(pure_store_draft(store, "other", &a) == PURE_OK); + CHECK(pure_draft_splice(b, 3, 0, (PureSpan){(const unsigned char *)"+", 1}) == PURE_OK); + CHECK(pure_draft_splice(state, 0, 0, (PureSpan){(const unsigned char *)"v", 1}) == PURE_OK); + CHECK(pure_draft_splice(a, 0, 0, (PureSpan){(const unsigned char *)"o", 1}) == PURE_OK); + PureDraft *batch[3] = {b, state, a}; + CHECK(pure_draft_commit_all(batch, 3, revisions) == PURE_OK && revisions[0] == 3 && revisions[1] == 1); + expect(path, "shared", &(Oracle){(unsigned char *)"two+", 4, 3}); + expect_draft(state, &(Oracle){(unsigned char *)"v", 1, 1}, scratch); + /* Other writers grow the file past compaction: it is replaced under the handle. */ + unsigned char *big = malloc(100000); + uint32_t seed = 5; + CHECK(big); + fill(big, 100000, &seed); + int64_t filler = 0; + char name[512]; + struct stat before, replaced; + snprintf(name, sizeof(name), "%s/pure.tree", path); + CHECK(stat(name, &before) == 0); + for (int i = 0; i < 8; ++i) CHECK(pure_store_put(path, "filler", filler, (PureSpan){big, 100000}, &filler) == PURE_OK); + CHECK(stat(name, &replaced) == 0 && replaced.st_ino != before.st_ino); + free(big); + PureDraft *after; + CHECK(pure_store_draft(store, "shared", &after) == PURE_OK); + expect_draft(after, &(Oracle){(unsigned char *)"two+", 4, 3}, scratch); + /* Drafts of the old file still read and commit (their extents are copied). */ + expect_draft(b, &(Oracle){(unsigned char *)"two+", 4, 3}, scratch); + CHECK(pure_draft_splice(b, 0, 0, (PureSpan){(const unsigned char *)">", 1}) == PURE_OK); + CHECK(pure_draft_commit(b, &revision) == PURE_OK && revision == 4); + expect(path, "shared", &(Oracle){(unsigned char *)">two+", 5, 4}); + pure_draft_close(after); + pure_draft_close(a); + pure_draft_close(b); + pure_draft_close(state); + pure_store_close(store); +} + +/* Threads of one process commit through their own handles at once: none is + * refused as busy, every commit lands, and edits of one shared key conflict + * instead of overwriting each other. */ +typedef struct { const char *path; unsigned number, conflicts; } Committer; + +static void *commit_many(void *context) { + Committer *c = context; + PureStore *store; + CHECK(pure_store_open(c->path, &store) == PURE_OK); + char key[16]; + snprintf(key, sizeof(key), "t%u", c->number); + for (unsigned i = 0; i < 40; ++i) { + PureDraft *own, *shared; + int64_t revisions[2]; + CHECK(pure_store_draft(store, key, &own) == PURE_OK && pure_store_draft(store, "shared", &shared) == PURE_OK); + CHECK(pure_draft_splice(own, pure_draft_size(own), 0, (PureSpan){(const unsigned char *)"x", 1}) == PURE_OK); + CHECK(pure_draft_splice(shared, pure_draft_size(shared), 0, (PureSpan){(const unsigned char *)"y", 1}) == PURE_OK); + PureDraft *batch[2] = {own, shared}; + PureResult result = pure_draft_commit_all(batch, 2, revisions); + CHECK(result == PURE_OK || result == PURE_CONFLICT); + if (result == PURE_CONFLICT) { + ++c->conflicts; + CHECK(pure_draft_commit(own, &revisions[0]) == PURE_OK); + } + pure_draft_close(own); + pure_draft_close(shared); + } + pure_store_close(store); + return NULL; +} + +static void threads(const char *path) { + pthread_t thread[4]; + Committer committer[4]; + for (unsigned i = 0; i < 4; ++i) { + committer[i] = (Committer){path, i, 0}; + CHECK(pthread_create(&thread[i], NULL, commit_many, &committer[i]) == 0); + } + unsigned conflicts = 0; + for (unsigned i = 0; i < 4; ++i) { + CHECK(pthread_join(thread[i], NULL) == 0); + conflicts += committer[i].conflicts; + } + PureStoredView view; + char key[16]; + for (unsigned i = 0; i < 4; ++i) { + snprintf(key, sizeof(key), "t%u", i); + CHECK(pure_store_get(path, key, PURE_OBJECT_LIMIT, &view) == PURE_OK && view.body.size == 40 && view.revision == 40); + pure_store_release(&view); + } + CHECK(pure_store_get(path, "shared", PURE_OBJECT_LIMIT, &view) == PURE_OK); + CHECK(view.body.size == 160 - conflicts && view.revision == (int64_t)(160 - conflicts)); + pure_store_release(&view); +} + +/* Commits that arrive while another is being published share the next + * publication: one head for all of them. */ +typedef struct { const char *path; unsigned number; PureResult result; } Single; + +static void *commit_one(void *context) { + Single *s = context; + PureStore *store; + PureDraft *draft; + char key[16]; + int64_t revision; + snprintf(key, sizeof(key), "g%u", s->number); + CHECK(pure_store_open(s->path, &store) == PURE_OK && pure_store_draft(store, key, &draft) == PURE_OK); + CHECK(pure_draft_splice(draft, 0, 0, (PureSpan){(const unsigned char *)key, strlen(key)}) == PURE_OK); + s->result = pure_draft_commit(draft, &revision); + CHECK(s->result != PURE_OK || revision == 1); + pure_draft_close(draft); + pure_store_close(store); + return NULL; +} + +static void grouped(const char *path) { + pthread_t thread[5]; + Single single[5]; + unsigned long before = pure_tree_test_publications(); + __atomic_store_n(&slow_once, 1, __ATOMIC_SEQ_CST); + for (unsigned i = 0; i < 5; ++i) { + single[i] = (Single){path, i, PURE_INVALID}; + CHECK(pthread_create(&thread[i], NULL, commit_one, &single[i]) == 0); + if (!i) pause_ms(20); /* the first is now publishing, slowly */ + } + for (unsigned i = 0; i < 5; ++i) CHECK(pthread_join(thread[i], NULL) == 0 && single[i].result == PURE_OK); + CHECK(pure_tree_test_publications() - before == 2); + char key[16]; + for (unsigned i = 0; i < 5; ++i) { + snprintf(key, sizeof(key), "g%u", i); + expect(path, key, &(Oracle){(unsigned char *)key, strlen(key), 1}); + } +} + +/* A crash during a commit's only sync can leave its head on the device + * without all of its bytes. A process that starts after it passes over that + * head to the previous commit, and the next writer overwrites both. */ +static void lost_commit(const char *path) { + unsigned char first[] = "first", second[] = "second"; + int64_t revision; + CHECK(pure_store_put(path, "lost", 0, (PureSpan){first, 5}, &revision) == PURE_OK); + Oracle o = {first, 5, revision}; + uint64_t start = published_end(path, "lost"); + CHECK(pure_store_put(path, "lost", o.revision, (PureSpan){second, 6}, &revision) == PURE_OK); + uint64_t end = published_end(path, "lost"); + char name[512]; + snprintf(name, sizeof(name), "%s/pure.tree", path); + int fd = open(name, O_RDWR); + off_t at = (off_t)(start + (end - start) / 2); + unsigned char byte; + CHECK(fd >= 0 && pread(fd, &byte, 1, at) == 1); + byte ^= 0xff; + CHECK(pwrite(fd, &byte, 1, at) == 1 && close(fd) == 0); + pure_tree_test_forget(); + expect(path, "lost", &o); + PureStore *store; + PureDraft *draft; + CHECK(pure_store_open(path, &store) == PURE_OK && pure_store_draft(store, "lost", &draft) == PURE_OK); + CHECK(pure_draft_revision(draft) == o.revision); + pure_draft_close(draft); + CHECK(pure_store_put(path, "lost", o.revision + 1, (PureSpan){first, 1}, &revision) == PURE_CONFLICT); + CHECK(pure_store_put(path, "lost", o.revision, (PureSpan){second, 3}, &revision) == PURE_OK); + o = (Oracle){second, 3, revision}; + CHECK(published_end(path, "lost") < end); /* it started where the lost commit did */ + CHECK(pure_store_draft(store, "lost", &draft) == PURE_OK && pure_draft_revision(draft) == o.revision); + pure_draft_close(draft); + pure_store_close(store); + pure_tree_test_forget(); + expect(path, "lost", &o); +} + +/* While a commit's sync is pending, this process keeps reading the previous + * version: a crash could still take the new one back. */ +static void *commit_held(void *context) { + int64_t revision; + CHECK(pure_store_put(context, "pending", 1, (PureSpan){(const unsigned char *)"new", 3}, &revision) == PURE_OK); + return NULL; +} + +static void pending_head(const char *path) { + int64_t revision; + CHECK(pure_store_put(path, "pending", 0, (PureSpan){(const unsigned char *)"old", 3}, &revision) == PURE_OK); + PureStore *store; + PureDraft *draft; + CHECK(pure_store_open(path, &store) == PURE_OK); + __atomic_store_n(&held_sync, 1, __ATOMIC_SEQ_CST); + pthread_t writer; + CHECK(pthread_create(&writer, NULL, commit_held, (void *)path) == 0); + while (__atomic_load_n(&held_sync, __ATOMIC_SEQ_CST) != 2) pause_ms(1); + CHECK(pure_store_draft(store, "pending", &draft) == PURE_OK && pure_draft_revision(draft) == 1); + pure_draft_close(draft); + expect(path, "pending", &(Oracle){(unsigned char *)"old", 3, 1}); + __atomic_store_n(&held_sync, 0, __ATOMIC_SEQ_CST); + CHECK(pthread_join(writer, NULL) == 0); + CHECK(pure_store_draft(store, "pending", &draft) == PURE_OK && pure_draft_revision(draft) == 2); + pure_draft_close(draft); + pure_store_close(store); + expect(path, "pending", &(Oracle){(unsigned char *)"new", 3, 2}); +} + +/* A commit costs one sync, and commits into the written zeros after the end + * leave the file size alone (no journal commit for a new size). */ +static void one_sync(const char *path) { + int64_t revision; + CHECK(pure_store_put(path, "sync", 0, (PureSpan){(const unsigned char *)"0", 1}, &revision) == PURE_OK); + uint64_t size = file_size(path); + unsigned long syncs = pure_tree_test_syncs(); + for (unsigned i = 0; i < 20; ++i) + CHECK(pure_store_put(path, "sync", revision, (PureSpan){(const unsigned char *)"x", 1}, &revision) == PURE_OK); + CHECK(pure_tree_test_syncs() - syncs == 20 && file_size(path) == size); +} + +static void clean(const char *path) { + DIR *entries = opendir(path); + CHECK(entries); + struct dirent *entry; + while ((entry = readdir(entries))) { + if (!strcmp(entry->d_name, ".") || !strcmp(entry->d_name, "..")) continue; + CHECK(strncmp(entry->d_name, ".pending-", 9)); + CHECK(unlinkat(dirfd(entries), entry->d_name, 0) == 0); + } + CHECK(closedir(entries) == 0 && rmdir(path) == 0); +} + +int main(void) { + void (*checks[])(const char *) = {oracle, amplification, faults, crashes, torn_head, + isolation, corruption, many_keys, revision_limit, keys, + drafts, typing, draft_conflict, draft_relocation, draft_faults, + draft_versions, anonymous_draft, batch, store_handle, threads, grouped, + lost_commit, pending_head, one_sync, records}; + for (size_t i = 0; i < sizeof(checks) / sizeof(*checks); ++i) { + char path[] = "/tmp/pure-tree-check-XXXXXX"; + CHECK(mkdtemp(path) && pure_store_init(path) == PURE_OK); + checks[i](path); + CHECK(pure_store_init(path) == PURE_OK); + clean(path); + } + puts("{\"tree_store\":\"random-oracle-balance-held-views-one-byte-amplification-faults-" + "process-crash-torn-head-lockless-isolation-compaction-corruption-1024-keys-revision-limit-key-listing-" + "draft-oracle-typing-arena-conflict-relocation-commit-faults-held-versions-anonymous-" + "atomic-batch-runtime-keys-store-handle-threads-group-commit-lost-commit-pending-head-one-sync-records\"}"); + return 0; +} |