mirror of
https://github.com/openai/codex.git
synced 2026-10-26 18:38:47 +00:00
Trust managed MITM CA in Linux sandboxes
This commit is contained in:
@@ -34,6 +34,7 @@ where
|
||||
P: AsRef<Path>,
|
||||
{
|
||||
let network_sandbox_policy = permission_profile.network_sandbox_policy();
|
||||
let mitm_ca_trust_bundle_path = network.and_then(NetworkProxy::mitm_ca_trust_bundle_path);
|
||||
let args = create_linux_sandbox_command_args_for_permission_profile(
|
||||
command,
|
||||
command_cwd.as_path(),
|
||||
@@ -41,6 +42,7 @@ where
|
||||
sandbox_policy_cwd,
|
||||
use_legacy_landlock,
|
||||
allow_network_for_proxy(/*enforce_managed_network*/ false),
|
||||
mitm_ca_trust_bundle_path.as_deref(),
|
||||
);
|
||||
let codex_linux_sandbox_exe = codex_linux_sandbox_exe.as_ref();
|
||||
// Preserve the helper alias when we already have it; otherwise force argv0
|
||||
|
||||
Reference in New Issue
Block a user