build(ci): fix OOM errors on macOS runner

[david-allison]

...by taking physical memory into account when deciding how many test processes to spawn.

Note

Assisted-by: Claude Opus 4.8 - diagnosis & some code

Purpose / Description

Even though macos-14 has 3 CPUs, there is only 7GB of memory.
Each test process has a max heap of 2GB, with the addition of Gradle this is enough to OOM the machine.

Fixes

• Fixes CI: OOM issues on macOS #21168

Approach

• Reserve 3GB of RAM for Gradle in our calculations
• Divide the remaining memory by 2GB per fork
• Cap the number of works

How Has This Been Tested?

CI-only

Learning (optional, can help others)

• macos-14 runner. 3 CPUs, 7GB RAM
• 3GB reserved for Gradle (org.gradle.jvmargs=-Xmx3072M) and the OS
• 4GB remaining
• Heap size: 2GB
• => 2 forks; previously 3

Checklist

• You have a descriptive commit message with a short title (first line, max 50 chars).
• You have commented your code, particularly in hard-to-understand areas
• You have performed a self-review of your own code
• UI changes: include screenshots of all affected screens (in particular showing any new or changed strings)
• UI Changes: You have tested your change using the Google Accessibility Scanner

[mikehardy]

I think the approach is solid - question on whether we can make the 3 non-magical

If answer is "nope", then merge at will

If answer is "yep", then after pulling the 3 dynamically, merge away with gusto

[mikehardy]

is there a way to get that 3GB value dynamically?

concern: we change the 3GB at some point to 4GB (it does creep up over time...) and now we still OOM even with 2 workers because the formula should be kicking out 1 vs 2 ?

[david-allison]

Thanks! I chose to parse -Xmx3072M"

Runtime.getRuntime().maxMemory() is dependent on GC flags

GC	maxMemory() for -Xmx3072M	/ GiB (floor)
G1 (JDK 21 default)	3221225472	3
ParallelGC	2863661056	2
SerialGC	3113877504	2

[david-allison]

@mikehardy Could I have an approve?

Request changes -> I believe I can only rebase directly, rather than use the merge queue

[mikehardy]

Runtime.getRuntime().maxMemory() is dependent on GC flags

Why can't anything be easy 😆

Gnarly code but also based on the flag-dependency of the easy way - an optimal solution

And so it goes - nifty fix, thank you!