Message ID | 20170228174001.23830-1-arnout@mind.be |
---|---|
State | Not Applicable |
Headers | show |
Hello, On Tue, 28 Feb 2017 18:39:54 +0100, Arnout Vandecappelle (Essensium/Mind) wrote: > Ever since we switched to gcc5 as default, a number of defconfigs have > been failing. Since we currently don't test the defconfigs this went > largely unnoticed. > > This series fixes the remaining defconfigs. Since I don't have access > to most of the boards, I couldn't test the changes, so I've opted for > the safe fix: apply the patch that adds gcc5 support to kernel and boot > loader. > > To avoid repeating the same exercise when we switch to gcc6 as default, > I've opted for the patch that removes the version-specific > compiler-gcc[345].h. > > Since the same kernel patch is used by a number of defconfigs, I've > put it in a common directory under board/, and use BR2_GLOBAL_PATCH_DIR > to refer to it. For U-Boot that wasn't possible because the context is > always slightly different. Therefore the patches are in board-specific > directories. The two Altera platforms can still share some of the > patches. > > This series has only been build-tested, not runtime tested. However, > the compiler-gcc.h change should be safe (if it breaks something, it > would be compilation itself because a definition is missing). > > Note: the yocto repository used by minnowboard_defconfig seems to be > either broken or unreliable. AFAIK minnowboard should be fully supported > by the upstream kernel, so it would be nice if someone who has the board > could try that. Thanks for all this work. However, it has already been noticed a while ago that these defconfigs were broken, I did a build test of all defconfigs some time ago, and reported the issue to the developers who contributed the defconfig originally. Peter also helped in this effort. See e-mails like: [Buildroot] Altera socdk defconfig no longer builds [Buildroot] Altera sockit defconfig no longer builds [Buildroot] via_imx6_vab820_defconfig fails to build etc. For all the boards you are fixing now, Peter or me contacted the original submitters. If they haven't been fixed by now, then I would instead suggest to simply drop support for those boards. Best regards, Thomas
>>>>> "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: Hi, > However, it has already been noticed a while ago that these defconfigs > were broken, I did a build test of all defconfigs some time ago, and > reported the issue to the developers who contributed the defconfig > originally. Peter also helped in this effort. See e-mails like: > [Buildroot] Altera socdk defconfig no longer builds > [Buildroot] Altera sockit defconfig no longer builds > [Buildroot] via_imx6_vab820_defconfig fails to build > etc. > For all the boards you are fixing now, Peter or me contacted the > original submitters. If they haven't been fixed by now, then I would > instead suggest to simply drop support for those boards. I agree. Especially as we don't have access to (most) of the boards, and older bootloaders/kernels sometimes do break at runtime with newer gcc versions. E.G. if the defconfigs aren't really supported and looked after, then we should pretend they are.
On 28-02-17 21:54, Peter Korsgaard wrote: >>>>>> "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: > > Hi, > > > However, it has already been noticed a while ago that these defconfigs > > were broken, I did a build test of all defconfigs some time ago, and > > reported the issue to the developers who contributed the defconfig > > originally. Peter also helped in this effort. See e-mails like: > > > [Buildroot] Altera socdk defconfig no longer builds > > [Buildroot] Altera sockit defconfig no longer builds > > [Buildroot] via_imx6_vab820_defconfig fails to build > > etc. > > > For all the boards you are fixing now, Peter or me contacted the > > original submitters. If they haven't been fixed by now, then I would > > instead suggest to simply drop support for those boards. > > I agree. Especially as we don't have access to (most) of the boards, and > older bootloaders/kernels sometimes do break at runtime with newer gcc > versions. > > E.G. if the defconfigs aren't really supported and looked after, then we > should pretend they are. ^^^^^^ shouldn't :-) I'll send removal patches tomorrow (hopefully). I've marked the series as rejected in patchwork. Regards, Arnout