20:15 <jaeger> Any reason we shouldn't go ahead and add compressed module support to kmod? gzip, xz, and zstd are all in core already

20:20 <beerman> I think the argument was mostly stability, but I'd agree with building the compression support

20:26 <jaeger> Is there something that suggests it would be unstable?

20:27 <beerman> not sure really

20:27 <jaeger> I'll give it a try on a system here for a bit and see if anything breaks

20:28 <jaeger> I don't foresee any issue but who knows?

20:28 <jaeger> Semi-related, what about adding zstd support to pkgutils?

20:29 <beerman> +1 from me as well

22:09 <farkuhar> It probably hasn't been observed yet in any real-world deployments of the 3.8 RC, but after commit 7bea2426d9 R_SERVICES gets defined as a one-element array when entering single-user mode. The correct quoting is in rc.shutdown, just copy that into rc.single for the fix.

22:32 <jaeger> funky

22:33 <farkuhar> ukky and I were discussing the idea to avoid bash arrays in rc.conf altogether, and just rely on whitespace-splitting of a string like "lo crond net". There are too many footguns in bash arrays; the extra syntax makes it easy for typos to slip in undetected.

22:42 <farkuhar> If you search for "array" in the ChangeLog https://tiswww.case.edu/php/chet/bash/CHANGES you'll notice a number of array-related bug fixes over the years. Those bugs related to users redefining IFS might caution us against thinking that "split multi-token string on whitespace" is a more robust solution.