* Re: sdk created with arago fido contains hardcoded paths to build system in setup script
2016-04-14 17:33 ` Denys Dmytriyenko
@ 2016-04-22 16:26 ` Thomas Kaufmann
2016-04-25 13:00 ` Thomas Kaufmann
1 sibling, 0 replies; 4+ messages in thread
From: Thomas Kaufmann @ 2016-04-22 16:26 UTC (permalink / raw)
To: Denys Dmytriyenko; +Cc: meta-arago@arago-project.org
Hi Denys
Thank you so much for this hint, I think it pointed me to the right direction.
I based an own sdk and toolchain recipe on meta-toolchain-arago*.
It took me some time to ge rid of the unneeded stuff, but now it bakes.
Also, I now base my images arago-base-minimal instead of core-image-minimal.
One issue remains, the sw_mainfest target() task fails, I think because of a non expanded shell variable.
In the log it looks for a file with ${image} in the name. So all varaibles except one are expanded in the call.
So for the time being I commented out most of sw_mainfest target() in tisdk_image.bbclass.
Another strange behavior is, about every second time I bitbake it parses linux-ti-staging, which I don't use and I did not find the relevant reference to it so far.
Regards Thomas
> -----Original Message-----
> From: Denys Dmytriyenko [mailto:denys@ti.com]
> Sent: Thursday, April 14, 2016 7:33 PM
> To: Thomas Kaufmann <Thomas.Kaufmann@duagon.com>
> Cc: meta-arago@arago-project.org
> Subject: Re: [meta-arago] sdk created with arago fido contains hardcoded
> paths to build system in setup script
>
> On Wed, Mar 23, 2016 at 12:12:08PM +0000, Thomas Kaufmann wrote:
> > Hi I tried to get an answer to this from the TI e2e community, but so
> > far without luck:
> >
> >
> https://e2e.ti.com/support/embedded/linux/f/354/p/500188/1811754#1811
> 7
> > 54
> >
> > My issue is this:
> >
> > When I create a SDK for my target platform using arago with the
> > populate_sdk command in bitbake. I receive a self extracting shell
> > script, which unpacks all the rootfs, compilers, libraries and
> > everything. Basically the SDK is working and I am able to develop for my
> target with it.
> >
> > One thing bothers me:
> >
> > in the environment-setup-cortexa8hf-vfp-non-linux-gnueabi file which
> > needs to be sourced before using the SDK, there are some hardcoded
> > paths which point to the linaro toolchain as it is installed on the
> > original system, where I actually build the SDK,
> >
> > e.g.: export CFLAGS="
> > -isystem/home/tkaufmann/gcc-linaro-4.9-2015.05-x86_64_arm-linux-
> gnueab
> > ihf/arm-linux-gnueabihf/include -fstack-protector -O2 -pipe -g
> > -feliminate-unused-debug-types"
> >
> > similar things are in CFLAGS, CXXFLAGS, LDFLAGS and CPPFLAGS
> >
> > The plan was to use the compiler and sysroot as it is available in the SDK.
> > This is working so far for sinple projects without separately
> > installing the linaro toolchain. Why are these linaro based paths in
> > the environment-setup script and how do I get rid of this to not confuse
> the users of the SDK?
> >
> > I appreciate your feedback
>
> Thomas,
>
> Sorry for the delay. I just checked our SDK and there are no hardcoded paths
> in there.
>
> Trying to figure out what's different I see that your setup script is called
> "environment-setup-cortexa8hf-vfp-non-linux-gnueabi" while ours is just
> simple "environment-setup". Also I see you mention populate_sdk in
> bitbake and now it becomes clear - you should be using meta-toolchain-
> arago-tisdk recipe where lots of stuff gets cleanup and configured properly...
>
> Attached is the current environment-setup script from our SDK.
>
> --
> Denys
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: sdk created with arago fido contains hardcoded paths to build system in setup script
2016-04-14 17:33 ` Denys Dmytriyenko
2016-04-22 16:26 ` Thomas Kaufmann
@ 2016-04-25 13:00 ` Thomas Kaufmann
1 sibling, 0 replies; 4+ messages in thread
From: Thomas Kaufmann @ 2016-04-25 13:00 UTC (permalink / raw)
To: Denys Dmytriyenko; +Cc: meta-arago@arago-project.org
Hi again
Just a quick followup. The problems with non expanded variables vanished, after I cleaned out my work directory and rebuilt everything.
The linux-ti-staging is still drawn into my build occasionally, I don't really understand why. But for now I think its fine.
Regards Thomas
> -----Original Message-----
> From: Thomas Kaufmann
> Sent: Friday, April 22, 2016 6:26 PM
> To: 'Denys Dmytriyenko' <denys@ti.com>
> Cc: meta-arago@arago-project.org
> Subject: RE: [meta-arago] sdk created with arago fido contains hardcoded
> paths to build system in setup script
>
> Hi Denys
>
> Thank you so much for this hint, I think it pointed me to the right direction.
> I based an own sdk and toolchain recipe on meta-toolchain-arago*.
> It took me some time to ge rid of the unneeded stuff, but now it bakes.
>
> Also, I now base my images arago-base-minimal instead of core-image-
> minimal.
>
> One issue remains, the sw_mainfest target() task fails, I think because of a
> non expanded shell variable.
> In the log it looks for a file with ${image} in the name. So all varaibles except
> one are expanded in the call.
> So for the time being I commented out most of sw_mainfest target() in
> tisdk_image.bbclass.
>
> Another strange behavior is, about every second time I bitbake it parses
> linux-ti-staging, which I don't use and I did not find the relevant reference to
> it so far.
>
> Regards Thomas
>
> > -----Original Message-----
> > From: Denys Dmytriyenko [mailto:denys@ti.com]
> > Sent: Thursday, April 14, 2016 7:33 PM
> > To: Thomas Kaufmann <Thomas.Kaufmann@duagon.com>
> > Cc: meta-arago@arago-project.org
> > Subject: Re: [meta-arago] sdk created with arago fido contains
> > hardcoded paths to build system in setup script
> >
> > On Wed, Mar 23, 2016 at 12:12:08PM +0000, Thomas Kaufmann wrote:
> > > Hi I tried to get an answer to this from the TI e2e community, but
> > > so far without luck:
> > >
> > >
> >
> https://e2e.ti.com/support/embedded/linux/f/354/p/500188/1811754#1811
> > 7
> > > 54
> > >
> > > My issue is this:
> > >
> > > When I create a SDK for my target platform using arago with the
> > > populate_sdk command in bitbake. I receive a self extracting shell
> > > script, which unpacks all the rootfs, compilers, libraries and
> > > everything. Basically the SDK is working and I am able to develop
> > > for my
> > target with it.
> > >
> > > One thing bothers me:
> > >
> > > in the environment-setup-cortexa8hf-vfp-non-linux-gnueabi file which
> > > needs to be sourced before using the SDK, there are some hardcoded
> > > paths which point to the linaro toolchain as it is installed on the
> > > original system, where I actually build the SDK,
> > >
> > > e.g.: export CFLAGS="
> > > -isystem/home/tkaufmann/gcc-linaro-4.9-2015.05-x86_64_arm-linux-
> > gnueab
> > > ihf/arm-linux-gnueabihf/include -fstack-protector -O2 -pipe -g
> > > -feliminate-unused-debug-types"
> > >
> > > similar things are in CFLAGS, CXXFLAGS, LDFLAGS and CPPFLAGS
> > >
> > > The plan was to use the compiler and sysroot as it is available in the SDK.
> > > This is working so far for sinple projects without separately
> > > installing the linaro toolchain. Why are these linaro based paths in
> > > the environment-setup script and how do I get rid of this to not
> > > confuse
> > the users of the SDK?
> > >
> > > I appreciate your feedback
> >
> > Thomas,
> >
> > Sorry for the delay. I just checked our SDK and there are no hardcoded
> > paths in there.
> >
> > Trying to figure out what's different I see that your setup script is
> > called "environment-setup-cortexa8hf-vfp-non-linux-gnueabi" while ours
> > is just simple "environment-setup". Also I see you mention
> > populate_sdk in bitbake and now it becomes clear - you should be using
> > meta-toolchain- arago-tisdk recipe where lots of stuff gets cleanup and
> configured properly...
> >
> > Attached is the current environment-setup script from our SDK.
> >
> > --
> > Denys
^ permalink raw reply [flat|nested] 4+ messages in thread