Hi Nobuhiro, all, On Wed, Sep 12, 2018 at 1:41 PM Baurzhan Ismagulov <[email protected]> wrote: > > On Wed, Sep 12, 2018 at 03:29:35PM +0900, Nobuhiro Iwamatsu wrote: > > > Are you saying that you want to use your Debian workstation to run > > > something > > > like buildroot-ng, to create a busybox-based root filesystem for a device? > > > > > > Or, are you wanting to create and then run a small-footprint Debian on a > > > device? > > > > I want to do both. > > I want to create and run a small-footprint Debian on a device. > > I agree with Bill, it would be easier to start from scratch. That said, the > patches from Crush and SLIND could be used as ideas. > > > > > As I recall it, Emdebian Crush ended because multiarch replaced much of > > > it, > > > and broke the rest of it. This was expected, since Crush was kind of a > > > workaround for pre-multiarch. > > > > About the cross compile environment, as you wrote, it became useless > > by support of M-A. > > I understand that there was a small rootfs using busybox and uclibc > > (This may not have been supported yet.) > > as another feature of Crush. Is this right? > > One footprint-related issue with SLIND were big Perl dependencies. Perl is > necessary for postinstall scripts. We've used perl-minimal and stripped > unnecessary dependencies. > > > > > There might be easier ways to get to where you want to be than to bring > > > Emdebian Crush back from the dead. > > > > https://wiki.debian.org/EmdebianCrush > > > > Indeed, when I saw this, as you wrote, I thought it would be better to > > consider another way than to back Crush. > > The list is valid if you want to make a generic distribution and do it right. > It is also the reason why Crush has been abandoned. That said, it's possible > to > bootstrap a minimal system for a specific product in finite amount of time and > keep improving it. If we would manage to upstream the changes to Debian in > form > of build options turned off by default and set up CI (ideally including the > cross-build stuff by Helmut & Co.), the effort might be somewhat more > manageable.
Would ELBE [1] solve your problem? [1] https://elbe-rfs.org/ Yegor
