Tag
Finding the actual cause when WordPress breaks.
Troubleshooting
The warning names the file and line that produced output too early, in the phrase in parentheses. Everything else in the message is the victim. Fix the file it names, and the login redirects, feeds and cookie errors that came with it stop together.
Troubleshooting
WordPress needs the mysqli extension and the PHP that is serving the site does not have it loaded. The database is fine. The extension is one checkbox in a hosting panel, one package on a server, or one line in a Docker image, and the error appears right after a PHP version change.
Troubleshooting
The file WordPress was asked to activate exists, but nothing in its first 8 KB says Plugin Name. That narrows it to a damaged header or an unreadable file; a missing file, a nested zip and a stray folder each produce a different message, or none.
Troubleshooting
The message ends with a question — is its parent directory writable by the server? — and the answer is usually no, but for one of two reasons: ownership on the folder, or a stale upload path from a previous server still stored in the database.
Troubleshooting
The browser is almost never the problem. WordPress sets a test cookie on the login form and looks for it on submit, and anything that changes the domain between those two requests — a URL mismatch, a cache, a proxy — makes the cookie vanish.
Troubleshooting
The zip you uploaded is not a theme, or the theme is inside it one level too deep. Marketplace downloads bundle the installable zip with documentation, and uploading the bundle is the cause nine times out of ten.
Troubleshooting
The limit is PHP's, not WordPress's, and raising it in wp-config.php cannot work because the directive is read before any PHP script runs. Where it can be raised depends on the server, and two other limits sit behind it.
Troubleshooting
WordPress tried to write a folder and the server said no. The message does not say which folder or why, and the three usual reasons — ownership, a full disk, a missing upgrade directory — each need a different fix.
Troubleshooting
WordPress checked a capability and the account did not have it. After a migration the usual reason is not the role at all — it is that the table prefix changed and the row describing the account's role no longer matches.
Troubleshooting
The number is specific and useful: 28 is always a timeout, never a refusal or a certificate problem. What matters next is which request timed out, because a loopback to your own site and a call to a payment provider need opposite fixes.
Troubleshooting
The installer refused because a directory with that name is already there, usually left by an install that failed partway. The folder is the evidence, and deleting the wrong one removes a working plugin.
Troubleshooting
The editor asked the REST API for JSON and got something else back. Whatever that something is — an HTML error page, a redirect, or a stray warning — is the actual fault, and it is visible in one browser tab.