Waiting for all clients to be closed before activating the new service worker is not usable, as it will lead to user keeping the old code.
Drop the logic for now.

Fetch the list of precache url dynamically in order to decouple the worker code from the list of files.

Calculate dynamically the list of files from a python script, depending on the web site configuration.

Update the service worker as soon as the web site modification date changes.
This (I hope) will simplify ERP5 upgrade handling.

See nexedi/erp5@2e3f7a8a

Disable the action, as the gadget is broken

This reverts commit 3bea5dc5.

Drop base tag, because it will probably break all erp5js forks, which hardcode url strings

This reverts commit edddfa32.

[erp5_web_renderjs_ui] Update service worker code. Client_id is null when it is the first request, in other words if request is navigate mode. Since major web browsers already implement client_id, if client_is is null, let's use the latest cache and don't get cache_key from CACHE_MAP and erp5js_cache.

[erp5_web_renderjs_ui] If service worker failed to install cache, unregister this service worker explicitly.

The name must be different per web site, else if the same service worker code is used by multiple web sites in ERP5 web site module, service worker does not work correctly.

[erp5_web_renderjs_ui] Fix translation script. Get service worker filename from layout property. Don't hardcode it.

[erp5_web_renderjs_ui] New client can use the latest cache without waiting for the new service worker to be activated.

And once client was associated with a cache, client keeps using the same cache.

To preserve the consistency of code and data, let the new service worker wait until all tabs and windows of the old version are closed.

Don't give request object itself to cache.match. Firefox's Cache Storage does not work properly when VARY contains Accept-Language. Give URL string instead, then cache.match works on both Firefox and Chrome.