On 12/5/25 16:19, enh wrote:
an alternative to TOYBOX_ON_ANDROID would be to just "default n" those
toys, and rely on the fact that Android has its own hand-maintained
.config file(s) anyway...

(but probably not worth doing that without having a way to drop
TOYBOX_FORK too.)

I'd generally like the build not to break after running config, which includes selinux and the library selections too.

In theory the library stuff could become a single "use external libraries if available" config option, and then __has_include(blah) for the specific ones. But that doesn't _quite_ cover SLL or the TOYBOX_LSM_NONE stuff we don't have internal implementations of...

TOYBOX_FORK already has a symbol to force the nommu codepaths for testing (to always use vfork() and marshall data via pipes and such, even on systems with mmu). Explicitly building ls without -Z is kinda similar (if we CAN build that way, then I want to be able to test it regardless of build environment), but I neither want feature micromanagement (which is why I've removed most command sub-options) and I don't want symbols to _disable_ stuff (config symbols are positive, not negative, "allnoconfig" not meaning allnoconfig is a pet peeve of mine with modern linux). So both LS_Z and LS_NO_Z are wrong as config symbols. :(

More missing design work. Dunno the answer yet.

Rob
_______________________________________________
Toybox mailing list
[email protected]
http://lists.landley.net/listinfo.cgi/toybox-landley.net

Reply via email to