IBX-12192: Skip SiteAccessStamp for uninitialized SiteAccess

[Steveb-p]

🎫 Issue	IBX-12192

Related PRs:

• IBX-12192: Added SiteAccess::MATCHING_TYPE_UNINITIALIZED constant core#796

Description:

SendMessageSiteAccessSubscriber only skipped stamping when SiteAccessServiceInterface::getCurrent() returned null. Turns out that's not what happens when a message gets dispatched from a context where no HTTP request ever matched a SiteAccess — e.g. a console command. getCurrent() instead returns the placeholder SiteAccess CoreBundle wires into the container before request matching (name default, matchingType = 'uninitialized'), so the subscriber stamped the envelope with that. On the consuming side, SiteAccessMiddleware then tried to resolve a SiteAccess literally named 'default' and blew up with NotFoundException: Could not find 'SiteAccess' with identifier 'default'.

This is what broke ibexa/data-intelligence-layer's CI, which dispatches messages from console-driven integration tests (failing run).

Fix mirrors the existing null check: skip stamping when $siteAccess->matchingType === SiteAccess::MATCHING_TYPE_UNINITIALIZED.

[mnocon]

I'm not sure I understand the description correctly, and just want to clarify:

I've always thought that when you don't specify the siteaccess with the command (the --siteaccess) option, the default siteaccess is used.

From the --siteaccess help description:

php bin/console cache:clear --help
#...
--siteaccess[=SITEACCESS]  SiteAccess to use for operations. If not provided, default siteaccess will be used

And by default siteaccess I mean the default_siteaccess setting:https://github.com/ibexa/recipes-dev/blob/master/ibexa/oss/4.6/config/packages/ibexa.yaml#L64

I didn't know that by omitting the option the siteaccess is not set at all (that's how I understand "uninitialized").
Can it affect the processing at all, compared to what would have happened if the default siteaccess was set?

I mean, can you get different results by running:

php bin/console app:my-command
php bin/console app:my-command --siteaccess=site

?

[mnocon]

Thinking about the following case:

Given the worker is run with:
php bin/console messenger:consume ibexa.messenger.transport --bus=ibexa.messenger.bus --siteaccess=site_B

When a command dispatching a message is invoked with:
php bin/console app:my-command

Result: SiteAccess is uninitialized -> no SiteAccessStamp, message is processed by the worker using the site_B context (default siteaccess for the worker)

When a command dispatching a message is invoked with:
php bin/console app:my-command --siteaccess=site

Result: SiteAccessStamp is attached, message is processed by the worker using the site siteaccess

[Steveb-p]

@mnocon I'll need to confirm this - there is some chance that default_siteaccess might work, and it wasn't working in tests in another package because there were missing dependencies / bundles.

But tbh it is very, very likely that what you are describing might also be bugged and default site access may not be set for the bin/console context.