On Fri, Aug 22, 2025 at 01:08:07PM +0200, Christian Brauner wrote: > On Thu, Aug 21, 2025 at 04:18:13PM -0400, Josef Bacik wrote: > > +/* > > + * As simple macro to define the inode state bits, __NAME will be the bit value > > + * (0, 1, 2, ...), and NAME will be the bit mask (1U << __NAME). The __NAME_SEQ > > + * is used to reset the sequence number so the next name gets the next bit value > > + * in the sequence. > > + */ > > +#define INODE_BIT(name) \ > > + __ ## name, \ > > + name = (1U << __ ## name), \ > > + __ ## name ## _SEQ = __ ## name > > I'm not sure if this is the future we want :D > I think it's harder to parse than what we have now. If it would be a generic sounding name like ENUM_BIT https://elixir.bootlin.com/linux/v6.16.2/source/fs/btrfs/misc.h#L16 one does not have to parse anything and the meaning is understandable in the context of enum. We've converted everything to that in btrfs and it's convenient to change just one line when adding or removing entries, not caring about the exact values.
