On Sat, 2022-02-05 at 15:54 +0100, Pali Rohár wrote: > On Saturday 05 February 2022 03:07:00 Marcel Ziswiler wrote: > > On Sat, 2022-02-05 at 01:54 +0100, Pali Rohár wrote: > > > On Saturday 05 February 2022 01:40:23 Marcel Ziswiler wrote: > > > > On Sat, 2022-02-05 at 01:25 +0100, Pali Rohár wrote: > > > > > $ kwboot -b /dev/ttyUSB0 > > > > > > > > Hm, at least kwboot from today's master does not allow -b without also > > > > giving it an image. > > > > > > This commit is part of master branch and added support for it: > > > https://source.denx.de/u-boot/u-boot/-/commit/c513fe47dca24de87a904ce7d71cfd8a390e1154 > > > > > > If it does not work then there is some bug which should be fixed. I have > > > tested it and it works on my setup. > > > > > > Can you check if you have that commit to ensure that we are not going to > > > test / debug older version? > > > > Yes, sure. I debugged it a little and I believe I found the issue. I guess > > the intention was for it to be > > run > > giving it a dash '-' as image argument for skipping. (Even though that > > usually would mean using stdin). > > Anyway, > > it fails as it then uses such dash as ttypath because optind did not get > > incremented. The following fixes > > it > > for me: > > > > diff --git a/tools/kwboot.c b/tools/kwboot.c > > index 2684f0e75a..d490c026e9 100644 > > --- a/tools/kwboot.c > > +++ b/tools/kwboot.c > > @@ -1768,7 +1768,8 @@ main(int argc, char **argv) > > if (prev_optind == optind) > > goto usage; > > if (argv[optind] && argv[optind][0] != '-') > > - imgpath = argv[optind++]; > > + imgpath = argv[optind]; > > + optind++; > > break; > > > > case 'D': > > > > Let me know if that is indeed the intention and I can send a proper fix for > > it. Maybe I can also update the > > documentation/usage in that respect. Thanks again. > > Now I see where is the issue. Your fix is incorrect because it breaks > other usage (e.g. specifying other options). > > The issue here is that argv[optind] is used also when it is the last > option on the command line -- which is not getopt() at all, as it is > parsed after the getopt() processing. > > So I think that correct fix should be something like this: > > bootmsg = kwboot_msg_boot; > if (prev_optind == optind) > goto usage; > - if (argv[optind] && argv[optind][0] != '-') > + if (optind < argc-1 && argv[optind] && > argv[optind][0] != '-') > imgpath = argv[optind++]; > break; > > > This should fix usage of all "kwboot -b /dev/tty", "kwboot -b -t /dev/tty" > and "kwboot -b image.kwb /dev/tty" usage. > > Could you test it?
Yes, this indeed works fine now. With that, and if we properly document such "standalone" usage, I would be fine with dropping tools/mrvl_uart.sh. So you can get my reviewed-by for this one. Reviewed-by: Marcel Ziswiler <mar...@ziswiler.com> And you get my tested-by for such patch fixing this as outlined above. Tested-by: Marcel Ziswiler <mar...@ziswiler.com> > Seems that I tested only -b -t combination (as I wanted to see that > bootrom correctly start sending NAK xmodem bytes). Yep, now I got you. Thanks!