On Thu, 11 Jun 2009 18:03:51 +0200 Wolfgang Denk <w...@denx.de> wrote:
> Dear Kim Phillips, > > In message <20090611105533.7aeec1ee.kim.phill...@freescale.com> you wrote: > > > > > The separator between options is ';', between sub-options is ','. > > > > > > So that would be 'dr_usb:mode=host,phy_type=ulpi; esdhc". > > ...and that's still what I like most of the differnet formats > discussed so far. > > > > if you don't want to type, things like this are possible but they > > have to depend on the order given: > > It's not only about not typing, it's also about being able to read it > easily. right, which it currently isn't, due to conflicting precedence levels for the different operator characters chosen. btw, there should be no severe loss of readability here: > > dr_usb.mode = host; .phy_type = ulpi; esdhc; ... > > however when automating/scripting concatenation of them, it's useful > > to not have to depend on their order: > > > > dr_usb.mode = host; esdhc; dr_usb.phy_type = ulpi; > > I see no reason that we need to support such a format. In C, you also > cannot mix initializations of fields from different structures :-) I view these as regular assignments, not initializations :) > > What do you think? > > 'dr_usb:mode=host,phy_type=ulpi; esdhc" still looks reasonable to me. the ':', which, in C, would be as though we're specifying a bit-field, and, in English, has higher precedence than ';' which makes it confusing because ';' is used elsewhere with higher precedence than the ':'. I'd almost prefer a space instead, making: "dr_usb mode=host, phy_type=ulpi; esdhc" but the dot operator '.' is the best candidate since it accurately represents /member/ access. just my 2c. Kim _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
