myxce/src/controller/testsuite/test_e2e_gc.c

/* P0-15d: GC E2E — retention + transient job cleanup
 *
 * Exercises cross-phase interactions:
 *   - GC retention predicates (P0-E) applied to concrete job/worker fixtures
 *   - JOB_STAT_TRANSIENT timeout behaviour
 *   - Checkpoint succeeds after WAL writes (P0-C integration smoke)
 */
#define _GNU_SOURCE
#include <dirent.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/stat.h>
#include <time.h>
#include <unistd.h>

#include <unity/unity.h>

#include "config.h"
#include "persist.h"
#include "xce/types.h"

/* ------------------------------------------------------------------ */
/* Fixtures                                                             */
/* ------------------------------------------------------------------ */

static job_t make_gc_job(unsigned int state, time_t finish_time,
                         uint64_t transient_since)
{
    job_t j;
    memset(&j, 0, sizeof(j));
    j.job.state           = state;
    j.job.finish_time     = finish_time;
    j.job.transient_since = (uint64_t)transient_since;
    return j;
}

static xce_worker_t make_gc_worker(uint32_t status, time_t dead_since)
{
    xce_worker_t w;
    memset(&w, 0, sizeof(w));
    w.status     = status;
    w.dead_since = dead_since;
    return w;
}

/* ------------------------------------------------------------------ */
/* Test 1: running job must be retained                                 */
/* ------------------------------------------------------------------ */

static void test_job_gc_retain_running_job(void)
{
    persist_config_t cfg = persist_config_defaults();
    job_t j = make_gc_job(JOB_STAT_RUN, 0, 0);
    TEST_ASSERT_TRUE(job_should_retain(&j, &cfg));
    printf("test_job_gc_retain_running_job: PASS\n");
}

/* ------------------------------------------------------------------ */
/* Test 2: expired finished job must not be retained                   */
/* ------------------------------------------------------------------ */

static void test_job_gc_cleanup_expired_finished_job(void)
{
    time_t now = time(NULL);
    persist_config_t cfg = persist_config_defaults();
    cfg.finished_retain_s = 3600;
    job_t j = make_gc_job(JOB_STAT_EXIT, now - 7200, 0);
    TEST_ASSERT_FALSE(job_should_retain(&j, &cfg));
    printf("test_job_gc_cleanup_expired_finished_job: PASS\n");
}

/* ------------------------------------------------------------------ */
/* Test 3: transient job times out                                     */
/* ------------------------------------------------------------------ */

static void test_transient_job_times_out(void)
{
    time_t now = time(NULL);
    persist_config_t cfg = persist_config_defaults();
    cfg.transient_job_timeout_s = 300;
    /* transient_since = now - 400 seconds — past the 300s window */
    job_t j = make_gc_job(JOB_STAT_TRANSIENT, 0, (uint64_t)(now - 400));
    TEST_ASSERT_FALSE(job_should_retain(&j, &cfg));
    printf("test_transient_job_times_out: PASS\n");
}

/* ------------------------------------------------------------------ */
/* Test 4: expired unavailable worker must not be retained             */
/* ------------------------------------------------------------------ */

static void test_worker_gc_cleanup_expired_unavailable(void)
{
    time_t now = time(NULL);
    persist_config_t cfg = persist_config_defaults();
    cfg.worker_retain_s = 86400; /* 24 h */
    xce_worker_t w = make_gc_worker(XCE_WORKER_UNAVAIL, now - 100000);
    TEST_ASSERT_FALSE(worker_should_retain(&w, &cfg));
    printf("test_worker_gc_cleanup_expired_unavailable: PASS\n");
}

/* ------------------------------------------------------------------ */
/* Test 5: checkpoint succeeds and creates snapshot (integration smoke)*/
/* ------------------------------------------------------------------ */

static void test_checkpoint_gc_removes_expired_entity(void)
{
    char tmpdir[256];
    strcpy(tmpdir, "/tmp/xce_e2e_gc_XXXXXX");
    TEST_ASSERT_NOT_NULL(mkdtemp(tmpdir));

    persist_t* p = calloc(1, sizeof(persist_t));
    TEST_ASSERT_NOT_NULL(p);

    persist_config_t cfg = persist_config_defaults();
    cfg.wal_sync_ms = 5000;
    cfg.snap_keep_n = 2;

    persist_set_workdir(p, tmpdir, "e2e-gc");
    TEST_ASSERT_EQUAL_INT(0, persist_init(p, &cfg, NULL));
    TEST_ASSERT_EQUAL_INT(0, persist_start(p));

    /* Write some WAL entries (representing state changes) */
    for (int i = 0; i < 5; i++)
        TEST_ASSERT_EQUAL_INT(0,
            persist_wal_async(p, WAL_JOB_UPDATE, "gc-test", 7));

    /* No live apply pump runs in this cluster-less fixture, so
     * last_applied_lsn stays 0 unless simulated -- persist_checkpoint now
     * stamps checkpoint_lsn from last_applied_lsn, not current_lsn
     * (checkpoint_lsn-vs-last_applied_lsn fix); simulate apply having
     * caught up to what was appended before the checkpoint runs. */
    atomic_store_explicit(&p->last_applied_lsn, p->current_lsn - 1,
                          memory_order_relaxed);

    /* Trigger checkpoint — must succeed and leave a snapshot */
    TEST_ASSERT_EQUAL_INT(0, persist_checkpoint(p));
    TEST_ASSERT_TRUE(p->checkpoint_lsn > 0);

    char snap_path[512];
    snprintf(snap_path, sizeof(snap_path), "%s/e2e-gc/data/snap", tmpdir);

    DIR* d = opendir(snap_path);
    TEST_ASSERT_NOT_NULL(d);
    int found = 0;
    struct dirent* e;
    while ((e = readdir(d)) != NULL) {
        size_t len = strlen(e->d_name);
        if (len > 6 && strcmp(e->d_name + len - 6, ".capnp") == 0)
            found++;
    }
    closedir(d);
    TEST_ASSERT_TRUE(found >= 1);

    persist_stop(p);
    persist_fini(p);
    free(p);
    printf("test_checkpoint_gc_removes_expired_entity: PASS\n");
}

/* ------------------------------------------------------------------ */
/* Suite entry point                                                    */
/* ------------------------------------------------------------------ */

void test_e2e_gc(void)
{
    UnitySetTestFile(__FILE__);
    printf("\n>> P0-15d: GC E2E — retention + transient job cleanup\n\n");

    RUN_TEST(test_job_gc_retain_running_job);
    RUN_TEST(test_job_gc_cleanup_expired_finished_job);
    RUN_TEST(test_transient_job_times_out);
    RUN_TEST(test_worker_gc_cleanup_expired_unavailable);
    RUN_TEST(test_checkpoint_gc_removes_expired_entity);
}