Incorrect warning for system requirements for APCu memory

Investigation here: https://drupal.slack.com/archives/C079NQPQUEN/p1754057444000909 that links to the original discussion in htmx where I found this.

Misplaced parents?

Bytes::toNumber(ini_get('apc.shm_size')) * ini_get('apc.shm_segments')

Don’t we want:

Bytes::toNumber(ini_get('apc.shm_size') * ini_get('apc.shm_segments'))

?

Opening an MR to try that

Bah, sorry. I hadn't looked at the ini_get() docs about apc settings. apc.shm_size indeed can come back as "32 MB" and friends, which is what we want to convert to a number before we multiply it by the number of segments.

It might be good to print what ini_get('apc.shm_size') and ini_get('apc.shm_segments') actually are now in the build env, and whether something isn't converting those values right in Bytes::toNumber.

@catch also suggested in slack that maybe we disable Status test for now to get builds working, then reinstate after finding the root cause.

Initial MR to skip the test for now.

Let's do this and unblock some builds.

Opened MR to check what the APC values were per #6, and it looks like `ini_get('apc.shm_segments')` might be empty, which would certainly be a problem.
https://git.drupalcode.org/issue/drupal-3539331/-/jobs/6083955

OK so apc.shm_segments has apparently been removed:
https://git.drupalcode.org/project/drupalci_environments/-/blob/dev/php/...
https://pecl.php.net/package/APCu/5.1.25

Committed/pushed to 11.x and cherry-picked to 11.2.x, 10.6.x and 10.5.x, thanks!

Don't usually like multiple commits from a single issue but given skipping the test is an emergency stop-gap let's leave this open and continue here maybe?

catch → closed merge request !12888

godotislate → closed merge request !12892

OK, used a ternary to set fallback multiplier of 1 if ini_get('apc.shm_segments') is empty. This passes tests in MR 12895.

Hah, bit of duplicated effort, since I had started down that same road.

However, it's not just ini_get('apc.shm_segments') that can be empty. So, too, can apcu_sma_info(TRUE)['num_seg']. So I moved the fallback check into a local var, and use that everywhere we check this, either directly or via apcu_sma_info().

dww → changed the visibility of the branch 3539331-fix-apcu-status-check to hidden.

Probably not critical anymore.