Somebody wants to add 4 more options to ls, which puts its flag count above 32, which means they need more optflags bits.
I've avoided making optflags a 64 bit entity before now because that sucks libgcc_s into the base for stuff like "echo". For a defconfig build it's not a big deal, and existing toys.optflags&FLAG_x things can theoretically still do a 32 bit mask (certainly on sane little endian platforms) and just _ignore_ the high bits. But because of the bloat to lib/args.c I'm tempted to add a CFG_TOYBOX_LONGFLAGS with a #define for the type change in toys.h, and then have commands that need it "select TOYBOX_LONGFLAGS" in their kconfig. Except... I didn't add a define for --longopt parsing because I prefer the code to be simple rather than micromanage it to be small in certain configurations when the defconfig build (which I _assume_ is the common case) is the same either way... So I'm probably just going to throw the switch and make optflags an unsigned long long now that a user's showed up who needs it. But I'm conflicted, and open to arguments against if you've got 'em... Rob _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
