[Buildroot] [PATCH 1/1] ti-sgx-*: bump from SDK 02.00 to SDK 04.00

Einar Jón tolvupostur at gmail.com
Thu May 31 20:39:45 UTC 2018


On 30 May 2018 at 16:49, Thomas Petazzoni <thomas.petazzoni at bootlin.com> wrote:
> Hello Einar,
>
> Please don't reply just to me, keep the mailing list in Cc. Thanks!
>
> On Wed, 30 May 2018 16:21:49 +0200, Einar Jón wrote:
>
>> > We definitely want a bisectable patch series, where everything works
>> > between each commit.
>>
>> I could do these a patchset:
>> Beaglebone_qt5 bump
>> package/ti-sgx-km bump
>> package/ti-sgx-um bump
>> package/ti-sgx-demos bump
>> new package ti-sgx-libgbm
>>
>> Is that preferred, or just the buildable:
>> ti-sgx and Beaglebone_qt5 bump
>> new package ti-sgx-libgbm
>
> This doesn't work because your ti-sgx-um package update has:
>
>         select BR2_PACKAGE_TI_SGX_LIBGBM
>
> which implies it requires ti-sgx-libgbm to exist first.
>
> Here it was you should do instead:
>
>         ti-sgx-libgbm: new package
>         ti-sgx-{um,km}: bump version
>         configs/beablebone_qt5: update this
>         configs/beablebone_qt5: update that

I've spent too much time with git - My order meant the git history
order, committing the new package first, and beaglebone last.

> Indeed, in beaglebone_qt5_defconfig, you're doing lots of changes that
> are completely unrelated to ti-sgx stuff: updating U-Boot, adding
> Qt5Wayland, adding am335x-bonegreen in the list of DTs, etc.

True. ti-sgx requires a 4.9 kernel, which beaglebone_defconfig has.
I think the only reason why beaglebone_qt5 was not bumped up to kernel
4.9 with beaglebone was that
the ti-sgx would have been broken. So bumping the other changes to
match seemed to make sense.
After this commit beaglebone_qt5_defconfig is bumped to the same level
as beaglebone_defconfig.
Qt5Wayland is added because wayland is a new dependency pulled in by libgbm.

>
>> From my committed package/ti-sgx-km/ti-sgx-km.mk, with line numbers
>>     22 ifeq ($(BR2_PACKAGE_TI_SGX_KM_AM335X),y)
>>     23 TI_SGX_KM_MAKE_ENV += TARGET_PRODUCT=ti335x
>>     24 else ifeq ($(BR2_PACKAGE_TI_KM_SGX_AM437X),y)
>>     25 TI_SGX_KM_MAKE_ENV += TARGET_PRODUCT=ti437x
>>     26 endif
>>
>>  From my committed package/ti-sgx-um/ti-sgx-um.mk
>>     19 ifeq ($(BR2_PACKAGE_TI_SGX_KM_AM335X),y)
>>     20 TI_SGX_UM_MAKE_ENV += TARGET_PRODUCT=ti335x
>>     21 else ifeq ($(BR2_PACKAGE_TI_SGX_KM_AM437X),y)
>>     22 TI_SGX_UM_MAKE_ENV += TARGET_PRODUCT=ti437x
>>     23 endif
>
> These lines don't exist in the ti-sgx-um package in the current master
> branch. So I don't see where the problem is.

Yep. these are only in my new commit.
In the current master, ti-sgx-km.mk uses these variables to choose a
target subfolder,
but ti-sgx-um had the same code for all target products. This has
changed in both repos,
and now they have only 3 targets, instead of 4 earlier.

>
> Just base your patches on the current master branch, and use the new
> name of the Config.in options.

OK. I'll send in a new patchset tomorrow.

-- 
Regards
Einar Jón

>> Perfectly compatible with busybox. I'm using it myself.
>> It's just a /bin/sh script with args "start"|stop|reload|restart,
>> calling modprobe/rmmod wrappers in local functions load_pvr and unload_pvr.
>> It does call  echo "Loaded PowerVR consumer services." instead of the standard
>> printf "Loading PowerVR consumer services:"  + OK/FAIL
>
> OK, fair enough then.
>
> Thanks,
>
> Thomas
> --
> Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons)
> Embedded Linux and Kernel engineering
> https://bootlin.com



More information about the buildroot mailing list