On Tue, 2014-10-14 at 15:41 +0100, Ian Campbell wrote:
> On Tue, 2014-10-14 at 02:34 +0100, Ben Hutchings wrote:
> > On Mon, 2014-10-13 at 16:30 +0100, Ben Hutchings wrote:
> > [...]
> > > Here are my generic changes, untested as I don't have such boards to
> > > test with. They aren't going to cover i2c-s3c2410 (I think) or
> > > USB-PHYs.
> > >
> > > For USB-PHYs, if MODULES=most then we would do:
> > > copy_modules_dir kernel/drivers/usb/phy
> > > or if MODULES=dep then:
> > > add_loaded_modules 'phy[-_]*'
> > [...]
> >
> > I've pushed these two changes to the git branch
> > benh/add-more-device-types. Please test whether this works for your
> > board.
>
> Upon install (on xgene) I get:
> /usr/sbin/mkinitramfs: 252: /usr/share/initramfs-tools/hook-functions: Syntax
> error: redirection unexpected
>
> Which is from the last line of add_loaded_modules:
> done < <(cat 2>/dev/null /lib/modules/$(uname -r)/modules.builtin)
>
> Removing what looked like a mistake extra < I get:
> /usr/sbin/mkinitramfs: 252: /usr/share/initramfs-tools/hook-functions: Syntax
> error: "(" unexpected
>
> Ah, as I suspected, checkbashisms is confirming that <() is a bashism.
D'oh.
> My fixup patch is below.
>
> With that in place it boots to a prompt on xgene for both modules=most
> and =dep.
>
> I need to regroove my arndale to test there, more info on that soon.
>
> Ian.
>
> diff --git a/hook-functions b/hook-functions
> index 7659517..30e57ed 100644
> --- a/hook-functions
> +++ b/hook-functions
> @@ -243,13 +243,13 @@ add_loaded_modules()
> manual_add_modules $(basename $module)
> fi
> done
> - while read module; do
> + cat 2>/dev/null /lib/modules/$(uname -r)/modules.builtin | while read
> module; do
> case "$module" in
> */$pattern.ko)
> manual_add_modules $(basename $module .ko)
> ;;
> esac
> - done < <(cat 2>/dev/null /lib/modules/$(uname -r)/modules.builtin)
> + done
> }
>
> # find and only copy root relevant modules
I try to avoid putting variable assignments in pipelines as they may or
may not run in a subshell (it varies between implementations). In this
case I think it's harmless but I'd rather avoid it anyway.
So I think I'll wrap this loop in a file test instead.
Thanks for testing.
Ben.
--
Ben Hutchings
Horngren's Observation:
Among economists, the real world is often a special case.
signature.asc
Description: This is a digitally signed message part
