On Thu, 2016-05-05 at 00:07 +0200, Lukas Wunner wrote:
> Hi Ben,
>
> I upgraded to tag debian/4.6_rc5-1_exp1 today and wanted to provide some
> quick feedback. First of all it's good to no longer have linux-tools
> separately, thanks for doing this.
>
>
> On first build I got messages like this because config.h is only generated
> later in the build and the target install-usbip failed:
>
> sed: can't read debian/build/build-tools/tools/usb/usbip/config.h: No such file or directory
>
> After invoking "make -f debian/rules binary-arch" a second time the package
> was built without a hitch because config.h was now there.
Ah, I see the problem. dpkg-buildpackage invokes the build target
before binary{,-arch,-indep}, so this isn't a problem for auto-builds.
> In rules.d/tools/lib/lockdep/Makefile, ln -s fails for liblockdep.so
> because the symlink already exists. Either this should be ln -sf or
> the command is superfluous.
It's not superfluous, and it normally works. You'll have to explain
what commands you're using that cause it to fail.
> About this issue I mentioned in my Linux 4.5 e-mail in February:
>
> On Thu, Feb 25, 2016 at 09:41:28PM +0000, Ben Hutchings wrote:
> >
> > On Thu, 2016-02-25 at 16:16 +0100, Lukas Wunner wrote:
> > >
> > > --- a/debian/patches/features/all/drivers-media-dvb-usb-af9005-request_firmware.patch
> > > +++ b/debian/patches/features/all/drivers-media-dvb-usb-af9005-request_firmware.patch
> > > @@ -13,11 +13,7 @@ a version of the script which is directly derived from the driver.
> > >
> > > --- a/drivers/media/usb/dvb-usb/Kconfig
> > > +++ b/drivers/media/usb/dvb-usb/Kconfig
> > > -@@ -227,10 +227,10 @@ config DVB_USB_OPERA1
> > > -
> > > - config DVB_USB_AF9005
> > > - tristate "Afatech AF9005 DVB-T USB1.1 support"
> > > -- depends on BROKEN
> > > +@@ -236,6 +236,7 @@ config DVB_USB_AF9005
> > > depends on DVB_USB
> > > select MEDIA_TUNER_MT2060 if MEDIA_SUBDRV_AUTOSELECT
> > > select MEDIA_TUNER_QT1010 if MEDIA_SUBDRV_AUTOSELECT
> > This file is also patched by debian/patches/debian/dfsg/drivers-media-
> > dvb-dvb-usb-af9005-disable.patch when building the 'orig' tarball.
> So this patch adds "depends on BROKEN":
> debian/patches/debian/dfsg/drivers-media-dvb-dvb-usb-af9005-disable.patch
>
> And this patch subsequently removes it:
> debian/patches/features/all/drivers-media-dvb-usb-af9005-request_firmware.patch
>
> Forgive me for being dense but what is the motivation for this?
The orig tarball should have minimal changes from the upstream source
for DFSG compliance. In the upstream source, the firmware for this
driver is built-in, so when we delete the firmware (done by
debian/patches/debian/dfsg/files-1) the driver can't be built. This
first patch, applied when building the orig tarball, documents that
breakage.
The second patch, applied when unpacking the source package, is a
bigger change that fixes the driver by adding the ability to load
firmware from a separate file.
> I keep bringing this up because it breaks my workflow: I use
> "quilt pop -a" to remove all debian patches before rebasing
> or switching branches, and later use "quilt push -a" when I
> actually want to compile. I can work around this by keeping
> a modified version of the second above-mentioned patch in
> my tree but I'm wondering if it might be possible to just not
> add the "depends on BROKEN" in the first place?
You seem to be asking us to make the orig tarball the same as the
upstream git tag.
In order to do that, we'll need to get those changes upstream in some
way. That probably will happen eventually, but it hasn't been a high
priority. Maybe you're the one to do that?
Ben.
--
Ben Hutchings
All the simple programs have been written, and all the good names taken.Attachment:
signature.asc
Description: This is a digitally signed message part