fix(proxy): only reload nginx-proxy when its config test passes

[mrrobot47]

Problem

reload_global_nginx_proxy() gated the proxy reload on if ( \EE::launch( '... nginx -t' ) ). But EE::launch() defaults to returning a ProcessRun object (not an exit code), and every object is truthy in PHP — so the if was always true and the config was regenerated and reloaded regardless of whether nginx -t passed. A broken nginx config (e.g. from a half-written cert pair or a malformed vhost rule) was therefore reloaded and persisted; the shared proxy then fails to start on its next restart, taking down every site behind it.

Compounding bug: nginx -t also ran before the docker-gen regeneration, so it validated the pre-regeneration config rather than the default.conf that actually gets served.

Fix

Regenerate default.conf first, then capture the nginx -t ProcessRun and gate on 0 === $test->return_code. On failure, warn with the test output and skip the reload (return false); on success, reload. This mirrors the working nginx -t && nginx -s reload pattern already used elsewhere in the codebase.

Notes

• Scope is this function's internal gate. The ~15 external callers that ignore the return value are unchanged; a skipped reload now returns false and emits an EE::warning, but wiring callers to react to that is a separate change.

Testing

Manual (needs the EE + Docker harness): introduce an invalid proxy config (bad cert/vhost), trigger a reload, and confirm it warns and does NOT reload/persist the broken config; with a valid config, confirm normal regenerate + reload.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.