Four states that look identical
| Symptom | Most likely stage | Next step |
|---|---|---|
| The update fails with an HTTP error | Fetch | Work from the status code |
| The update reports success, but the profile size is unusually small | Empty body, login page or error text | Check the content type of what was returned |
| The card looks normal, the proxy page is empty | Not activated, validation or load failure | Switch the active profile and read the log |
| Only some nodes are missing | Override, script, filter, or an upstream change | Disable custom processing as a contrast |
Before acting, record the card's update time, whether it shows an error, which profile is active, and whether the proxy page is completely blank or merely missing entries.
Step 1: did you actually download a profile
- Confirm the address came from the provider panel's subscription or one-click import entry, not the marketing site or a login page.
- Inspect the response type in a safe environment. Do not hand the full address to an online converter or public checker.
- If the body is an HTML login page, a captcha, an error message or blank, the client can complete the request and still have no proxies.
- If the plan expired, the quota ran out or the token was reset, regenerate it in the panel. No client setting recovers that.
The URL itself normally carries a credential. If it leaked, rotate it in the panel — deleting a screenshot is not rotation.
Step 2: is the profile active, or merely listed
- Select the newly updated profile on the profiles page.
- Confirm it is set as the active configuration, not just stored in the list.
- Open the proxy page and look at the group names. Old names mean the new configuration has not been applied.
- Save the switch time, then read the log for validation failure, apply failure or a core that did not start.
Do not delete the previous profile. Keeping one that loaded before is what lets you tell a bad subscription apart from a client that cannot load anything.
Step 3: disable overrides, scripts and filters
Clash Verge Rev can modify a downloaded profile through merge rules, scripts and other overrides. A bad proxy-group reference, a failing script or a node-name filter can leave the source subscription fully populated while the final configuration references nothing.
- Back up the current settings and local customisations.
- Temporarily disable the overrides, scripts and filters associated with this profile.
- Update and activate again, and observe only whether nodes return.
- Re-enable one item at a time to find which one empties the list.
Do not paste a stranger's complete override over your configuration, and do not disable certificate verification to silence an error.
Step 4: let the log separate a failed load from a stale UI
Keep the first error, not the last few dozen lines. A YAML parse error, a field-type error, a proxy-group reference error or failed to apply config means the configuration never reached the core — fix the content or the override. If the log shows the profile applied and networking works while the proxy page stays blank, quit and relaunch normally before concluding anything about the interface.
Version-specific reports of “update succeeded but nodes do not display” exist in the project's issues. A single report proves a symptom occurred in one environment, not that every empty list is a client defect. Finish the contrasts above first, then search the issue tracker for your exact version.
Step 5: main config and proxy-providers are two fetches
Mihomo's proxy-providers documentation defines a remote provider as its own object with url, path, interval and health checks. The main configuration can load while the provider file still fails to download, fails to be written, or fails to parse. “Subscription update succeeded” describes one step of the main profile.
| Main config | Provider | Next step |
|---|---|---|
Contains direct proxies with a non-zero count | Not involved | Compare the active profile with the runtime config after overrides |
Mostly references proxy-providers | Provider update shows an HTTP or fetch error | Record the status and earliest error; check access conditions with the source |
| — | Provider fetched, log says permission denied | Check the application data directory permissions; do not run the client elevated as a permanent fix |
| — | Provider has nodes, proxy page still empty | The group may not reference the provider, or a filter removed it: back up, disable overrides, check each group's use and filter |
Clash Verge Rev issue #4144 describes a macOS user on 2.3.2 where “update all” reported success while updating the provider alone returned 503, and the log also showed a remote fetch EOF and a local permission denied. That shows two failure stages coexisting in one case; it does not mean every 503 comes from file permissions. Read your own first error and the specific provider before deleting files because a dialog said success.
Safe recovery order
- Turn TUN and the system proxy off, and confirm plain networking returns.
- Switch back to the most recent profile that loaded correctly.
- Keep the error time, app version, core version and a masked log.
- Remove the new profile only after confirming it cannot load. Do not clear all subscriptions and application data at once.
If the system proxy works and only TUN breaks connectivity, that is a capture, routing or DNS question — not this page's empty-list problem.
When it belongs to the provider or the project
- Multiple compatible clients return an empty list for the same subscription: ask the provider to check the content, plan and token.
- The raw profile contains proxies and validation still fails after overrides are disabled: report to the client project with a masked error.
- Only one version fails while an older build or another device works: record reproducible steps and search the official issues. Do not install third-party repacks.
What to record
- Whether the update reported success, and the profile size if the client shows it.
- Which profile is active, and whether the proxy page is empty or partial.
- The first error in the log, with its timestamp.
- Whether the profile uses direct
proxiesorproxy-providers. - Whether disabling overrides restored the nodes, and which item did it.
- Client version, core version and OS.