High memory usage after updating to NJS 0.8.6

[xbb123]

Describe the bug

The NJS 0.8.6 has much higher memory footprint than 0.8.5. On our stage server, when running our NJS test cases, the previous version 0.8.5 use about 0.8% of memory. Running the same test case set, the memory usage will go up to 25% and easily gets OOM killed when running pressure tests. Referring to my previous issue: https://github.com/nginx/njs/issues/185 We have discovered that variables (lists) with high numbers of elements will consume memory quite quickly; although this may not being the only factor: removing array with long variable does not reduce the memory footprint to normal level. Downgrading to NJS 0.8.5 resolved the issue.

To reproduce

Steps to reproduce the behavior: Referring to my previous issue: https://github.com/nginx/njs/issues/185 for gist

Your NGINX logs here

Oct  8 12:53:59 Stage kernel: [1096221.064717] oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=systemd-networkd.service,mems_allowed=0,global_oom,task_memcg=/system.slice/nginx.service,task=nginx,pid=1536276,uid=0
Oct  8 12:53:59 Stage kernel: [1096221.064737] Out of memory: Killed process 1536276 (nginx) total-vm:2945532kB, anon-rss:1742516kB, file-rss:2304kB, shmem-rss:0kB, UID:0 pgtables:5028kB oom_score_adj:0
Oct  8 12:53:59 Stage kernel: [1096221.448365] systemd[1]: snapd.service: Watchdog timeout (limit 5min)!
Oct  8 12:53:59 Stage kernel: [1096221.453954] systemd[1]: snapd.service: Killing process 1433291 (snapd) with signal SIGABRT.
Oct  8 12:54:00 Stage kernel: [1096221.779352] systemd[1]: nginx.service: Control process exited, code=exited, status=1/FAILURE
Oct  8 12:54:00 Stage kernel: [1096221.851480] systemd[1]: session-59088.scope: Deactivated successfully.
Oct  8 12:54:00 Stage kernel: [1096221.856485] systemd[1]: session-59088.scope: Consumed 18.556s CPU time.
Oct  8 12:54:00 Stage kernel: [1096222.067511] systemd[1]: snapd.service: Main process exited, code=exited, status=2/INVALIDARGUMENT
Oct  8 12:54:00 Stage kernel: [1096222.067523] systemd[1]: snapd.service: Failed with result 'watchdog'.
Oct  8 12:54:00 Stage kernel: [1096222.068074] systemd[1]: snapd.service: Consumed 16.648s CPU time.
Oct  8 12:54:00 Stage kernel: [1096222.206459] systemd[1]: snapd.service: Scheduled restart job, restart counter is at 2.
Oct  8 12:54:00 Stage kernel: [1096222.206683] systemd[1]: Stopped Snap Daemon.
Oct  8 12:54:00 Stage kernel: [1096222.206736] systemd[1]: snapd.service: Consumed 16.648s CPU time.
Oct  8 12:54:00 Stage kernel: [1096222.217663] systemd[1]: Starting Snap Daemon...
Oct  8 12:54:01 Stage kernel: [1096223.505220] systemd[1]: systemd-journald.service: Main process exited, code=dumped, status=6/ABRT
Oct  8 12:54:01 Stage kernel: [1096223.505323] systemd[1]: systemd-journald.service: Failed with result 'watchdog'.
Oct  8 12:54:01 Stage kernel: [1096223.506197] systemd[1]: systemd-journald.service: Consumed 4min 50.232s CPU time.
Oct  8 12:54:01 Stage kernel: [1096223.506913] systemd[1]: systemd-journald.service: Scheduled restart job, restart counter is at 1.
Oct  8 12:54:01 Stage kernel: [1096223.507648] systemd[1]: Stopped Journal Service.
Oct  8 12:54:01 Stage kernel: [1096223.507720] systemd[1]: systemd-journald.service: Consumed 4min 50.232s CPU time.
Oct  8 12:54:01 Stage kernel: [1096223.514640] systemd[1]: Starting Journal Service...
Oct  8 12:54:02 Stage kernel: [1096223.587813] systemd-journald[1536807]: File /var/log/journal/b3a5fdbd1f9c4041acc272a500182f33/system.journal corrupted or uncleanly shut down, renaming and replacing.
Oct  8 12:54:02 Stage kernel: [1096223.783435] systemd[1]: Started Journal Service.
Oct  8 12:54:02 Stage systemd[1]: nginx.service: A process of this unit has been killed by the OOM killer.
Oct  8 12:54:02 Stage systemd[1]: nginx.service: Main process exited, code=killed, status=9/KILL
Oct  8 12:54:02 Stage sh[1536724]: kill: (1536276): No such process
Oct  8 12:54:02 Stage systemd[1]: systemd-journald.service: Watchdog timeout (limit 3min)!
Oct  8 12:54:02 Stage systemd[1]: systemd-journald.service: Killing process 309 (systemd-journal) with signal SIGABRT.

### Expected behavior

A clear and concise description of what you expected to happen.

### Your environment

- Version of njs or specific commit 0.8.6
- Version of NGINX if applicable 1.27.2
- List of other enabled nginx modules if applicable
- OS: Ubuntu 24.04 & 22.04 exactly the same behavior 

### Additional context

Thank you!

[xeioex]

Hi @xbb123,

I believe I found the relevant issue. Could you, please, test that it fixes your problem? See https://github.com/nginx/njs/pull/798.