772919 - Release all OfflineCache custom profile files after: download failure/update failure/no update

Status: RESOLVED WONTFIX

(Core :: Networking: Cache, defect)

Description

[Honza Bambas (:mayhemer)]

+++ This bug was initially created as a clone of Bug #760067 +++

Bug 760067 implements shutdown of the offline cache device after a successful download or update.

I've ripped out the part that shuts the offline cache device down after a failure (when the new cache version is discarded) because:

- it conflicts with the patch from bug 769291 and it is not that simple to have a stable patch quickly before the train leaves
- it is not strictly necessary now because preloads will always be expected to succeed before a web app is started and we don't have any update-periodically feature right now

Comment 1

[Bill Walker [:bwalker] [@wfwalker]]

Per WebRT bug triage, not considered a blocker

Comment 2

[Alex Keybl [:akeybl]]

Is there any fallout outside of WebRT? If not, if WebRT bug triage doesn't feel this is a feature blocker, we don't feel this is something we should track for release. Holding on a final decision though.

Comment 3

[Honza Bambas (:mayhemer)]

(In reply to Alex Keybl [:akeybl] from comment #2)
> Is there any fallout outside of WebRT? If not, if WebRT bug triage doesn't
> feel this is a feature blocker, we don't feel this is something we should
> track for release. Holding on a final decision though.

This is my POV of a developer here:  a patch for this bug can be relatively complicated, so it should spend some time on m-c first.  I don't want to rush during next 2 days to have a stable patch for this anyway.

Fallout out of WebRT and web applications is zero.  The custom profile update is used only to pre-cache web apps.  And when a web app is successfully pre-cached, we release all files (that is already fixed in bug 760067).  When a web app is not pre-cached, then it also shouldn't be possible to start it (AFAICT).  Also, I believe this bug is not highly critical, since sqlite is capable of sharing files open by multiple process (I do this relatively often with sqlite3 shell).  Then, after the update is done (either successful or not), we don't manipulate the cache from the host (Firefox) process any more.

Comment 4

[Alex Keybl [:akeybl]]

Thanks Honza - given that, I agree there's no reason to track for release.