0.1.10

0.1.10

• If another WordPress install shares your database and its table prefix starts with yours (wp_shop_ beside wp_), its tables are no longer in your backup, where restoring your site would have replaced them. The backup lists them as left out.
• Restoring over your own site works when its tables have foreign keys, such as WooCommerce's download log. The restored tables reference each other, and each key keeps its name. Over your own site the name returns when you delete the copy kept from before the restore, which holds it until then.
• Replacing the site address after a restore to a new address is safe to resume on MyISAM tables. A slice the host stopped part way through a batch replaced some rows twice when the new address held the old one (a.com to a.com/blog).
• A restore runs only the statements the plugin's dump writes, and refuses a backup that names a file outside the site.
• Triggers on your tables are listed as left out. The backup does not hold them.
• A backup SafeGrd could not record, because it could not be reached, is recorded with the next backup.
• A slice that runs past its time limit no longer frees the lock of the slice that took over, so two slices never run at once. A restore cannot begin while a backup slice starts, and Disconnect waits up to 30 seconds for a running slice.
• The copy a restore keeps aside is not served by Apache or LiteSpeed.
• A storage link that expires during a restore is signed again.
• Delete the copy says so when the database refuses, instead of reporting it deleted.

0.1.9

• Back up before a plugin, theme or core update: a setting, off until you turn it on, under Tools, SafeGrd, Settings, or wp safegrd schedule --before-updates=on. The update waits up to 20 seconds for the backup in wp-admin and for the whole backup from WP-CLI, and goes ahead if the backup fails. Recent backups, the restore list and wp safegrd snapshots label such a backup "Before update: to ".

0.1.8

• Restored files are written through the WordPress filesystem API. Only a file larger than the memory a request may use, and the download archive, are written as a stream.

0.1.7

• Downloads are written under uploads/safegrd-downloads, not in wp-content, and the backup leaves that directory out.
• A download or a restored file the disk could not take in full now fails with "the disk may be full", instead of being offered for download or swapped into the site short.
• The Connect card asks SafeGrd nothing until the site is connected.

0.1.6

• Restore single plugins, themes or tables from a backup. Every other plugin, theme and table stays as it is. The picker lists each with the version in the backup and the one installed now.
• Download single tables as a SQL dump, or single plugins or themes as a tar, with wp safegrd download --tables, --plugins or --themes.
• wp safegrd contents lists the plugins, themes and tables a backup holds.
• The Updates, Plugins and Themes screens show how old the last backup is, with Back up now, while updates wait.
• Restore & download, redone: this site's backups first, with a filter for the others. Open one to see each part's size, the WordPress and PHP versions, and its plugins, themes and tables. A review lists what will be replaced before anything starts.
• Put the copy back undoes the last restore, and pressing it again restores it again. wp safegrd restore --undo does the same.
• Copies kept by earlier restores are listed with their size, with Delete them, or wp safegrd restore --delete-older.
• A restore of some plugins, themes or tables no longer moves the next backup a week out.
• A failed restore no longer hides the copy of the last one that worked.
• Recent backups show each backup's id and Open, and say Test-restored or Backed up.
• Logs mark a run that stopped without finishing as stopped.
• wp safegrd snapshots --site and --format=json, wp safegrd logs, and wp safegrd status shows the last test restore.
• Reading what a backup holds takes a third of the requests it did.
• A backup or restore the host stops three times at the same point fails, saying what it was doing, instead of starting again for good.
• A restore stopped while swapping the restored site in is finished by the next slice, and one that fails there can be put back. Before, the site stayed half swapped and backups were refused.
• A restore stopped while loading a table loads it again from its start, instead of failing with "Duplicate entry".
• A whole-plugins restore keeps this plugin in the plugins directory throughout.
• The maintenance page a stopped swap leaves lasts a minute, not ten.
• A file that shrinks while it is read is kept as read, not padded with zeros.
• A restore checks the disk has room for the files it writes beside the site before it starts, as a full disk would take the running site down.
• Restores read and write their list of files hundreds of rows at a time, and backups look up a directory's files in one query.

0.1.5

• A restore or a download of a site with thousands of files waits when SafeGrd asks it to slow down, instead of stopping with "Rate limit exceeded".