Bugzilla

(In reply to Jan Varga [:janv] from comment #19)
> We need to solve this in general somehow, but in this concrete case we should just throw away entire cache ?

I think the general plan is:
1) Enhance mozStorage in bug 1125157 to provide corruption notifications.
2) On corruption, wipe the storage bucket, which currently means the entire origin.  (Currently this is specified by https://storage.spec.whatwg.org/#storage-endpoints to include LocalStorage but we can do a carve-out.)  I think we've discussed having the DirectoryLock class expose a method for this which could then notify the nsIClearDataService.

Doing a localize removal of the cache is probably an okay stop-gap since the ServiceWorkerManager/ScriptLoader will notice the storage has gone away the next time it goes to try and wake up the ServiceWorker and will remove the registration.

(In reply to Jan Varga [:janv] from comment #19)
> We need to solve this in general somehow, but in this concrete case we should just throw away entire cache ?

I think the general plan is:
1) Enhance mozStorage in bug 1125157 to provide corruption notifications.
2) On corruption, wipe the storage bucket, which currently means the entire origin.  (Currently this is specified by https://storage.spec.whatwg.org/#storage-endpoints to include LocalStorage but we can do a carve-out.)  I think we've discussed having the DirectoryLock class expose a method for this which could then notify the nsIClearDataService.

Doing a localized removal of the cache (database) without doing things at the bucket/origin level is probably an okay stop-gap since the ServiceWorkerManager/ScriptLoader will notice the storage has gone away the next time it goes to try and wake up the ServiceWorker and will remove the registration.