Good Reasons for Removing Content from a Fossil Repository
Fossil is designed to keep all historical content forever. Fossil purposely makes it difficult for users to delete content. Old content is part of the project's *ahem* fossil record and should be maintained indefinitely to maintain an accurate history of the project.
Nevertheless, there may occasionally arise legitimate reasons for deleting content. Such reasons include:
- Spammers inserted inappropriate content into a wiki page, forum post, or ticket. Fossil lets you easily hide or amend such content, but since it is not a legitimate part of the project's history, there is no value in keeping it, so it is best removed permanently.
- A file that contains trade secrets or that is under someone else's copyright was accidentally committed and needs to be backed out.
- A malformed control artifact was inserted and is disrupting the operation of Fossil.
- A legitimate legal request was received requiring content to be removed. This would most likely be related to the accidental intellectual property error or spam cases listed above. Some countries recognise software patents, and so allow legal claims targetting code commits. Some countries can require publicly-available encryption software to be taken down if it is committed to the DAG without the correct government authorisation.
All of these are rare cases: Fossil is designed to foil spammers up front, legally problematic check-ins should range from rare to nonexistent, and you have to go way out of your way to force Fossil to insert bad control artifacts. Therefore, before we get to methods of permanently deleting content from a Fossil repos, let's give some alternatives that usually suffice, which don't damage the project's fossil record:
When a forum post or wiki article is "deleted," what actually happens is that a new empty version is added to the Fossil repository. The web interface interprets this as "deleted," but the prior version remains available if you go digging for it.
When you close a ticket, it's marked in a way that causes it to not show up in the normal ticket reports. You usually want to give it a Resolution such as "Rejected" when this happens, plus possibly a comment explaining why you're closing it. This is all new information added to the ticket, not deletion.
When you fossil rm a file, a new manifest is checked into the repository with the same file list as for the prior version minus the "removed" file. The file is still present in the repository; it just isn't part of that version forward on that branch.
If you make a bad check-in, you can shunt it off to the side by amending it to put it on a different branch, then continuing development on the prior branch:
$ fossil amend abcd1234 --branch BOGUS --hide
$ fossil up trunk
The first command moves check-in ID abcd1234 (and any subsequent check-ins on that branch!) to a branch called BOGUS, then hides it so it doesn't show up on the timeline. You can call this branch anything you like, and you can re-use the same name as many times as you like. No content is actually deleted: it's just shunted off to the side and hidden away. You might find it easier to do this from the Fossil web UI in the "edit" function for a check-in.
The second command returns to the last good check-in on that branch so you can continue work from that point.
When the check-in you want to remove is followed by good check-ins on the same branch, you can't use the previous method, because it will move the good check-ins, too. The solution is:
$ fossil merge --backout abcd1234
That creates a diff in the check-out directory that backs out the bad check-in abcd1234. You then fix up any merge conflicts, build, test, etc., then check the reverting change into the repository. Again, nothing is actually deleted; you're just adding more information to the repository which corrects a prior check-in.
Exception: Non-versioned Content
It is normal and expected to delete data which is not versioned, such as usernames and passwords in the user table. The fossil scrub command will remove all sensitive non-versioned data from a repository.
The scrub command will remove user 'bertina', along with their password, any supplied IP address, any concealed email address etc. However, in the DAG, commits by 'bertina' will continue to be visible unchanged even though there is no longer any such user in Fossil.
Fossil provides a mechanism called "shunning" for removing content from a repository.
Every Fossil repository maintains a list of the hash names of "shunned" artifacts. Fossil will refuse to push or pull any shunned artifact. Furthermore, all shunned artifacts (but not the shunning list itself) are removed from the repository whenever the repository is reconstructed using the "rebuild" command.
Shunning lists are local state
The shunning list is part of the local state of a Fossil repository. In other words, shunning does not propagate to a remote repository using the normal "sync" mechanism. An artifact can be shunned from one repository but be allowed to exist in another. The fact that the shunning list does not propagate is a security feature. If the shunning list propagated then a malicious user (or a bug in the fossil code) might introduce a shun record that would propagate through all repositories in a network and permanently destroy vital information. By refusing to propagate the shunning list, Fossil ensures that no remote user will ever be able to remove information from your personal repositories without your permission.
The shunning list does not propagate to a remote repository by the normal "sync" mechanism, but it is still possible to copy shuns from one repository to another using the "configuration" command:
fossil configuration pull shun remote-url
fossil configuration push shun remote-url
The two command above will pull or push shunning lists from or to the remote-url indicated and merge the lists on the receiving end. "Admin" privilege on the remote server is required in order to push a shun list. In contrast, the shunning list will be automatically received by default as part of a normal client "pull" operation unless disabled by the "auto-shun" setting.
Note that the shunning list remains in the repository even after the shunned artifact has been removed. This is to prevent the artifact from being reintroduced into the repository the next time it syncs with another repository that has not shunned the artifact.
Managing the shunning list
The complete shunning list for a repository can be viewed by a user with "admin" privilege on the "/shun" URL of the web interface to Fossil. That URL is accessible under the "Admin" button on the default menu bar. Items can be added to or removed from the shunning list. "Sync" operations are inhibited as soon as the artifact is added to the shunning list, but the content of the artifact is not actually removed from the repository until the next time the repository is rebuilt.
When viewing individual artifacts with the web interface, "admin" users will usually see a "Shun" option in the submenu that will take them directly to the shunning page and enable that artifact to be shunned with a single additional mouse click.