| Message ID | 20240624181714.1293865-1-rs@ti.com |
|---|---|
| Headers | show |
| Series | gcc 14 and bsp provider fixup | expand |
On Mon, Jun 24, 2024 at 01:17:12PM -0500, rs@ti.com wrote: > From: Randolph Sapp <rs@ti.com> > > Resolve some issues afer the new bsp-provider selection logic was added and add > some patches to help with compilation flag checks in the KM. FWIW, the new bsp-provider selection logic was not causing any new issues. Anyone using meta-ti master and scarthgap from early February until late May were already exposed to this. While 6.1 LTS was still the default, this change to 23.3 got merged on Feb 9: https://git.yoctoproject.org/meta-ti/commit/?id=737807fc4360b7fe3cbe493a6104558c4658a046 Separate new 24.1 DDK for the 6.6 LTS was added on May 2, still disabled: https://git.yoctoproject.org/meta-ti/commit/?id=d89907ec2322f23271917d75f052217dd8a2c27a And then 6.6 LTS got flipped to become the default on May 23: https://git.yoctoproject.org/meta-ti/commit/?id=d41d49cfdc6deee33a40ab44d453167c36462c89 When adding the BSP provider code and setting preferred versions, I did see the first commit that switched branches from 6.1 to 6.6, and I was considering reverting it (like you are proposing here with patch #1). But I didn't want to upset anyone and figured since this was already the default config for almost 4 months (!) of using the updated 23.3 DDK with 6.1 LTS, I just left it alone. So, the BSP provider logic just formalized all the existing defaults. > This includes equivalent patches from the following post for GCC 14 compat: > https://lists.yoctoproject.org/g/meta-ti/topic/106806264 > > Randolph Sapp (2): > ti-img-rogue-*_23.3: gcc 14 compat and branch sel > ti-img-rogue-driver_24.1: gcc 14 compat > > .../powervr-drivers/ti-img-rogue-driver_23.3.6512818.bb | 4 ++-- > .../powervr-drivers/ti-img-rogue-driver_24.1.6554834.bb | 2 +- > .../powervr-umlibs/ti-img-rogue-umlibs_23.3.6512818.bb | 2 +- > 3 files changed, 4 insertions(+), 4 deletions(-) > > -- > 2.45.1 >
On Mon Jun 24, 2024 at 3:06 PM CDT, Denys Dmytriyenko wrote: > On Mon, Jun 24, 2024 at 01:17:12PM -0500, rs@ti.com wrote: > > From: Randolph Sapp <rs@ti.com> > > > > Resolve some issues afer the new bsp-provider selection logic was added and add > > some patches to help with compilation flag checks in the KM. > > FWIW, the new bsp-provider selection logic was not causing any new issues. Fair, a change to existing behavior technically isn't a bug. That being said, the driver was rendered non-functional on AM62 as a result. Not a bug to you, but still a bug to me. > Anyone using meta-ti master and scarthgap from early February until late May > were already exposed to this. > > While 6.1 LTS was still the default, this change to 23.3 got merged on Feb 9: > https://git.yoctoproject.org/meta-ti/commit/?id=737807fc4360b7fe3cbe493a6104558c4658a046 > > Separate new 24.1 DDK for the 6.6 LTS was added on May 2, still disabled: > https://git.yoctoproject.org/meta-ti/commit/?id=d89907ec2322f23271917d75f052217dd8a2c27a > > And then 6.6 LTS got flipped to become the default on May 23: > https://git.yoctoproject.org/meta-ti/commit/?id=d41d49cfdc6deee33a40ab44d453167c36462c89 > > When adding the BSP provider code and setting preferred versions, I did see > the first commit that switched branches from 6.1 to 6.6, and I was considering > reverting it (like you are proposing here with patch #1). But I didn't want to > upset anyone and figured since this was already the default config for almost > 4 months (!) of using the updated 23.3 DDK with 6.1 LTS, I just left it alone. > > So, the BSP provider logic just formalized all the existing defaults. The dt breaking changes made to kernel 6.6 came after I was required to get a build of the module up and running on kernel 6.6. External modules are the first things blamed during a kernel migration. For a little while that branch, despite being tagged 6.6, was actually cross compatible with 6.1 and 6.6. I did not intend for people to use it on 6.1, nor did I intend for it to be a long term solution, assuming that when 6.6 was selected as the new default the need for cross compatibility could be dropped and our proposed dt change from last year could finally be introduced... That assumption didn't play out as a new decision was made and this bsp logic was added post branch shuffling. Truly, there are no easy solutions. - Randolph > > This includes equivalent patches from the following post for GCC 14 compat: > > https://lists.yoctoproject.org/g/meta-ti/topic/106806264 > > > > Randolph Sapp (2): > > ti-img-rogue-*_23.3: gcc 14 compat and branch sel > > ti-img-rogue-driver_24.1: gcc 14 compat > > > > .../powervr-drivers/ti-img-rogue-driver_23.3.6512818.bb | 4 ++-- > > .../powervr-drivers/ti-img-rogue-driver_24.1.6554834.bb | 2 +- > > .../powervr-umlibs/ti-img-rogue-umlibs_23.3.6512818.bb | 2 +- > > 3 files changed, 4 insertions(+), 4 deletions(-) > > > > -- > > 2.45.1 > >
From: Randolph Sapp <rs@ti.com> Resolve some issues afer the new bsp-provider selection logic was added and add some patches to help with compilation flag checks in the KM. This includes equivalent patches from the following post for GCC 14 compat: https://lists.yoctoproject.org/g/meta-ti/topic/106806264 Randolph Sapp (2): ti-img-rogue-*_23.3: gcc 14 compat and branch sel ti-img-rogue-driver_24.1: gcc 14 compat .../powervr-drivers/ti-img-rogue-driver_23.3.6512818.bb | 4 ++-- .../powervr-drivers/ti-img-rogue-driver_24.1.6554834.bb | 2 +- .../powervr-umlibs/ti-img-rogue-umlibs_23.3.6512818.bb | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-)