Siteferry 0.6.0

First public release.

Added

• New admin screens: a home screen with the operations and this site's state, a pull/push wizard (connect, choose, review, transfer), a progress screen with per-folder bars and a measured speed, a result screen with what changed, undo and the backup, and a connection screen with a masked key and permission switches.
• The step loop keeps running while you move between Siteferry screens, with a floating button back to the operation.
• File review for pulls and pushes: the comparison shows new, changed and destination-only files before anything moves, and you can bring only new files. WP-CLI: --only-new.
• A pull or push pauses instead of failing when the other site stops answering, and resumes from where it stopped. WP-CLI retries a paused job 5 times, every 30 seconds.
• While files transfer, the progress shows files and megabytes for the folder in transit, and each folder logs a line when it finishes.
• IBM Plex Sans is bundled (SIL OFL 1.1): the admin page loads nothing from other servers.
• Spanish (es_ES) translation. The admin interface is translatable, and the script uses wp.i18n.
• uninstall.php: deleting the plugin removes its options, temporary tables and working folder.
• GitHub Actions CI: PHPCS, and PHPUnit on PHP 8.1 and 8.3 with MySQL 8.0 and MariaDB 10.6.
• WordPress.org readme.txt, the GPL-2.0-or-later license file, READMEs in English and Spanish, and the Harbor brand (logo, icon and banners).

Changed

• Source code, comments, WP-CLI help and log messages are in English.
• Admin assets are now assets/siteferry.js and assets/siteferry.css.
• Confirmations and messages use the admin page's own dialog instead of the browser's alert and confirm boxes.

Fixed

• When the two sites' clocks differ by more than 5 minutes (for example after the computer slept), the request is signed again once with the other site's time instead of failing.
• A pull now finishes in the request that swaps the tables, so the folder swaps no longer fail with "cookie check failed" once this site's users are replaced. A lost session gets a notice that says to log in again and resume.
• A push to a site whose web server limits the request size (HTTP 413, such as nginx's client_max_body_size) sends smaller requests instead of failing. If even the smallest is refused, it pauses and says which limit to raise.
• A PHP notice printed on the other site (with display_errors on) no longer breaks a pull or push with "Invalid response from the remote site (HTTP 200)". That error now also shows the start of what the other site answered.
• An incoming push no longer replaces a paused job or one waiting for its review on the receiving site.
• A pushed chunk cut off on the receiving site after part of it was applied is completed instead of applied twice.
• Importing a file larger than one upload chunk failed on the second chunk.
• Repeating an import step after a cut-off request failed with "Duplicate entry", or silently duplicated rows in tables without a primary key. The load now resumes where it stopped.
• The same problem in copy, pull, export and backup steps, which could duplicate rows or dump statements. Also fixed: a repeated finalize step failed after the tables had already been swapped.