mirror of
https://github.com/DeusData/codebase-memory-mcp.git
synced 2026-11-04 18:58:00 +00:00
test(repro): refresh #363 memory-axis repro for shipped + clamped CBM_MEM_BUDGET_MB
The #363 memory axis (CBM_MEM_BUDGET_MB) has shipped, so this reproduce-first case is now a permanent GREEN guard, not an open bug. Its header claimed the override "does not exist" and asserted the budget is honoured "regardless of host RAM" — both stale: the resolver now clamps the override to effective (cgroup/host) RAM. On a <4 GiB host the old 4096 MiB assertion would clamp and falsely flip the bug board RED. Rewrite the header to describe the shipped + clamped behaviour and drop the override to 128 MiB, which is at/below any realistic host's RAM so it is honoured exactly and stays stable on every runner. The clamp path itself is unit-tested in tests/test_mem.c (resolve_budget_override_clamped_to_total). Signed-off-by: Sam Li <yangsec888@gmail.com>
This commit is contained in:
Sam Li committed
1 parent
04f95ce7b8
commit
cbdd60b1e5
1 file changed
+30
-78
@@ -1,74 +1,38 @@
|
||||
/*
|
||||
* repro_issue363.c — Reproduce-first case for OPEN bug #363.
|
||||
* repro_issue363.c — Regression guard for bug #363 (both axes now FIXED).
|
||||
*
|
||||
* Issue: #363 — "Linux: cbm_system_info / cbm_default_worker_count don't
|
||||
* respect cgroup CPU/memory limits"
|
||||
*
|
||||
* ROOT CAUSE (two distinct axes):
|
||||
* Both axes of #363 have shipped; this case is now a permanent GREEN guard:
|
||||
*
|
||||
* CPU axis — FIXED in v0.8.0 (commit a5a3d1d).
|
||||
* cbm_detect_cgroup_cpus() reads /sys/fs/cgroup/cpu.max (v2) or
|
||||
* .../cpu/cpu.cfs_quota_us + .../cpu/cpu.cfs_period_us (v1) and the
|
||||
* result is used by detect_system_linux() in system_info.c:226.
|
||||
* cbm_default_worker_count() also honours the CBM_WORKERS env override
|
||||
* (commit d952238). Both are thoroughly tested in test_platform.c.
|
||||
* .../cpu/cpu.cfs_quota_us + .../cpu/cpu.cfs_period_us (v1); the result
|
||||
* feeds detect_system_linux(). cbm_default_worker_count() also honours the
|
||||
* CBM_WORKERS env override (commit d952238). Both tested in test_platform.c.
|
||||
*
|
||||
* Memory axis — STILL OPEN (confirmed by reporter @mayurpise in the last
|
||||
* open comment on #363, 2026-06-25).
|
||||
* cbm_detect_cgroup_mem() similarly reads /sys/fs/cgroup/memory.max (v2)
|
||||
* or .../memory/memory.limit_in_bytes (v1), and detect_system_linux()
|
||||
* uses it (system_info.c:229). BUT: there is NO env-override knob on
|
||||
* the memory axis. The CPU axis has CBM_WORKERS; the memory side has
|
||||
* nothing. On a bare-metal host with no enclosing cgroup, users cannot
|
||||
* cap cbm_mem_init's budget without wrapping the process in a cgroup
|
||||
* scope (as @mayurpise's workaround shows).
|
||||
* Memory axis — FIXED.
|
||||
* cbm_detect_cgroup_mem() reads /sys/fs/cgroup/memory.max (v2) or
|
||||
* .../memory/memory.limit_in_bytes (v1); detect_system_linux() applies
|
||||
* min(cgroup, host). The missing env knob — the "EXACT OPEN GAP" this
|
||||
* repro was filed for — now exists: cbm_mem_init() reads CBM_MEM_BUDGET_MB
|
||||
* and routes it through the pure cbm_mem_resolve_budget() (explicit
|
||||
* override > implicit detection, mirroring CBM_WORKERS).
|
||||
*
|
||||
* EXACT OPEN GAP:
|
||||
* A CBM_MEM_BUDGET_MB environment variable (analogous to CBM_WORKERS) that
|
||||
* cbm_mem_init() checks before computing g_budget from info.total_ram.
|
||||
* If set to a valid integer N, cbm_mem_init() should set
|
||||
* g_budget = N * 1024 * 1024, honouring it regardless of cgroup or host RAM.
|
||||
*
|
||||
* WHY THIS TEST IS RED:
|
||||
* cbm_mem_init() (src/foundation/mem.c) reads cbm_system_info().total_ram
|
||||
* and multiplies by ram_fraction. It does NOT call cbm_safe_getenv for
|
||||
* CBM_MEM_BUDGET_MB — the override path does not exist. Setting
|
||||
* CBM_MEM_BUDGET_MB=4096 has no effect; cbm_mem_budget() returns a value
|
||||
* derived from host RAM (or cgroup RAM when inside a container), not from
|
||||
* the env var. The assertion ASSERT_EQ(cbm_mem_budget(), 4096*1024*1024)
|
||||
* therefore fails on any host whose cgroup or physical RAM != exactly 4 GiB.
|
||||
*
|
||||
* ROOT CAUSE LOCATION:
|
||||
* src/foundation/mem.c, cbm_mem_init(), after the mimalloc option block
|
||||
* (currently around line 126):
|
||||
* cbm_system_info_t info = cbm_system_info();
|
||||
* g_budget = (size_t)((double)info.total_ram * ram_fraction);
|
||||
* The fix is to insert a cbm_safe_getenv("CBM_MEM_BUDGET_MB", ...) lookup
|
||||
* BEFORE this line and, if valid, set g_budget directly without involving
|
||||
* info.total_ram — mirroring the CBM_WORKERS pattern in
|
||||
* cbm_default_worker_count() (system_info.c:290).
|
||||
*
|
||||
* INTENDED FIX:
|
||||
* 1. In cbm_mem_init(): read CBM_MEM_BUDGET_MB; if set to a valid positive
|
||||
* integer, use that value (in bytes) as g_budget and log it.
|
||||
* 2. Test: set CBM_MEM_BUDGET_MB=4096, call cbm_mem_init(0.5), assert
|
||||
* cbm_mem_budget() == 4096 * 1024 * 1024. This test goes GREEN when
|
||||
* the override is wired.
|
||||
* 3. Complementary: on Linux, confirm cbm_system_info().total_ram is capped
|
||||
* by the cgroup memory limit when present — already covered in
|
||||
* test_platform.c via cbm_detect_cgroup_mem() unit tests, but an
|
||||
* integration path via cbm_system_info() is untestable without a seam
|
||||
* that lets callers override the hardcoded "/sys/fs/cgroup" root in
|
||||
* detect_system_linux() (system_info.c:229).
|
||||
* IMPORTANT — clamp semantics:
|
||||
* cbm_mem_resolve_budget() clamps the override to effective (cgroup/host) RAM
|
||||
* so it can never claim more than the box has. So a *large* override is NOT
|
||||
* honoured verbatim on a small host — it clamps to total_ram. This guard
|
||||
* therefore uses a small budget (REPRO363_BUDGET_MB) that is at/below any
|
||||
* realistic host's RAM, so it is honoured exactly and the assertion is stable
|
||||
* on every runner. The clamp behaviour itself is unit-tested separately in
|
||||
* tests/test_mem.c (resolve_budget_override_clamped_to_total).
|
||||
*
|
||||
* NOTE on cbm_mem_init() caching:
|
||||
* g_budget is initialised once via atomic_compare_exchange_strong.
|
||||
* The test must run in a process where cbm_mem_init() has NOT been called
|
||||
* yet, OR the test must reset g_initialized — neither is supported today.
|
||||
* The repro works as written because the repro runner does not call
|
||||
* cbm_mem_init() before this suite. If the initialisation guard is an
|
||||
* issue, the fix also needs a cbm_mem_reset_for_test() hook (test-only,
|
||||
* guarded by CBM_TEST_HOOKS or similar).
|
||||
* g_budget is initialised once via atomic_compare_exchange_strong, so this
|
||||
* guard relies on running before any suite that calls cbm_mem_init(); the
|
||||
* repro runner does not init the budget before this suite.
|
||||
*/
|
||||
|
||||
#include "test_framework.h"
|
||||
@@ -77,27 +41,20 @@
|
||||
#include <stdint.h>
|
||||
#include <stdlib.h>
|
||||
|
||||
#define REPRO363_BUDGET_MB 4096UL
|
||||
/* Deliberately small so the resolver honours it exactly (never clamps) on any
|
||||
* host — see the clamp-semantics note above. */
|
||||
#define REPRO363_BUDGET_MB 128UL
|
||||
#define REPRO363_BUDGET_BYTES (REPRO363_BUDGET_MB * 1024UL * 1024UL)
|
||||
|
||||
/*
|
||||
* repro_issue363_mem_budget_env_override
|
||||
*
|
||||
* Precondition: CBM_MEM_BUDGET_MB=4096 is set before cbm_mem_init() is
|
||||
* called. The budget should be 4096 MiB regardless of host RAM or cgroup.
|
||||
*
|
||||
* RED condition (current code):
|
||||
* cbm_mem_init() ignores CBM_MEM_BUDGET_MB entirely; cbm_mem_budget()
|
||||
* returns host-RAM * fraction, not 4 GiB. The assertion fires unless the
|
||||
* test runner happens to be on a machine whose effective RAM is exactly
|
||||
* 8 GiB with fraction=0.5 — essentially never.
|
||||
*
|
||||
* GREEN condition (after fix):
|
||||
* cbm_mem_init() reads CBM_MEM_BUDGET_MB, finds "4096", sets
|
||||
* g_budget = 4096 * 1024 * 1024. The assertion passes on any machine.
|
||||
* Precondition: CBM_MEM_BUDGET_MB is set before the process's first
|
||||
* cbm_mem_init(). The budget must equal the override (honoured exactly, since
|
||||
* it is below any host's effective RAM), proving the env knob is wired.
|
||||
*/
|
||||
TEST(repro_issue363_mem_budget_env_override) {
|
||||
cbm_setenv("CBM_MEM_BUDGET_MB", "4096", 1);
|
||||
cbm_setenv("CBM_MEM_BUDGET_MB", "128", 1);
|
||||
|
||||
cbm_mem_init(0.5);
|
||||
|
||||
@@ -105,11 +62,6 @@ TEST(repro_issue363_mem_budget_env_override) {
|
||||
|
||||
cbm_unsetenv("CBM_MEM_BUDGET_MB");
|
||||
|
||||
/*
|
||||
* RED on current code: budget derives from host/cgroup RAM, not the env
|
||||
* var. On any machine where effective RAM != 8192 MiB this fails.
|
||||
* GREEN once CBM_MEM_BUDGET_MB is wired in cbm_mem_init().
|
||||
*/
|
||||
ASSERT_EQ((long long)budget, (long long)REPRO363_BUDGET_BYTES);
|
||||
|
||||
PASS();
|
||||
|
||||
Reference in new issue
Block a user