GHSA-59fj-m2j6-hcxh

Glances: `--disable-config-exec` does not cover on-alert action commands (incomplete fix of CVE-2026-53925)

## Summary In Glances 4.5.5 the `--disable-config-exec` flag was extended (GHSA-3vwc-qwhc-3mj7) to stop `secure_popen()` from interpreting the shell operators `&&`, `|` and `>` in **AMP** command values taken from the configuration file. The hardening was not applied to the **on-alert action** command path, which reads its command lines from the same configuration file. As a result, with `--disable-config-exec` enabled, a configured alert action that contains `>` (file redirection), `&&` (chaining) or `|` (pipe) still has those operators interpreted, allowing arbitrary file write / command chaining at the privilege of the glances process when the alert triggers. ## Affected code `glances/actions.py` (Glances 4.5.5, latest): ```python ret = secure_popen(cmd_full) # line 111 — no allow_operators=, defaults to True ``` By contrast the AMP modules were fixed: ```python # glances/amps/default/__init__.py:69 self.set_result(secure_popen(res, allow_operators=self.allow_operators()).rstrip()) # glances/amps/systemv/__init__.py:60 res = secure_popen(self.get('service_cmd'), allow_operators=self.allow_operators()) ``` ## PoC (benign) `glances.conf`: ```ini [cpu] user_critical=1 user_critical_action=echo MARKER > /tmp/poc_marker ``` Run `glances --disable-config-exec` and generate CPU load. When the cpu `user` alert reaches CRITICAL, `/tmp/poc_marker` is created — i.e. the `>` operator was interpreted despite `--disable-config-exec`. The same `>` in an `[amp_*]` `command` value is correctly *not* interpreted. ## Impact Arbitrary file write (`>`), command chaining (`&&`) and pipe (`|`) from config-defined alert actions, contrary to the guarantee of `--disable-config-exec`. Trust boundary = the glances configuration file. ## Suggested fix Pass `allow_operators=not args.disable_config_exec` from `GlancesActions.run()` into `secure_popen()` (GlancesActions already holds `args`). ## Credit Reported via responsible-disclosure incomplete-fix measurement study.