On 2018年06月20日 07:18, Qu Wenruo wrote: > > > On 2018年06月19日 22:47, David Sterba wrote: >> On Mon, Jun 18, 2018 at 09:05:59PM +0800, Qu Wenruo wrote: >>>> New code needs to be tested, documented and maintained, that's the cost >>>> I find too high for something that's convenience for people who are used >>>> to some shell builtins. >>> >>> The biggest problem is, the behavior isn't even consistent across >>> btrfs-progs. >>> mkfs.btrfs accept such out-of-order parameters while btrfs not. >>> >>> And most common tools, like commands provided by coretutil, they don't >>> care about the order. >>> The only daily exception is 'scp', which I found it pretty unhandy. >>> >>> And just as Paul and Hugo, I think there are quite some users preferring >>> out-of-order parameter/options. >> >> Because of the feedback and interest of the relaxed mixed >> option/argument syntax, I don't object anymore. > > And in fact, the behavior of btrfs is caused by @optind wrongly initialized. > The fix is just one line (along some comments). > > https://patchwork.kernel.org/patch/10473173/ > > So no or little pressure on maintenance.
I'm totally wrong. Since there are cases we call check_argc_*() directly using @optind, reinitialize it to 0 would cause them to fail. I'll update the patch to manually reset optind to 0 only for cases we call getopt(). And indeed, it adds extra maintenance effort. Thanks, Qu > > Thanks, > Qu >
signature.asc
Description: OpenPGP digital signature