On Thu, Jul 27, 2023 at 08:34:44AM +0000, Sahin, Okan wrote:
> >On Tue, Jul 18, 2023 at 08:55:02AM -0700, Nathan Chancellor wrote:
> >
> ><snip>
> >
> >> > +static struct regulator_desc max77857_regulator_desc = {
> >> > + .ops = &max77857_regulator_ops,
> >> > + .name = "max77857",
> >> > + .linear_ranges = max77857_lin_ranges,
> >> > + .n_linear_ranges = ARRAY_SIZE(max77857_lin_ranges),
> >> > + .vsel_mask = 0xFF,
> >> > + .vsel_reg = MAX77857_REG_CONT2,
> >> > + .ramp_delay_table = max77857_ramp_table[0],
> >> > + .n_ramp_values = ARRAY_SIZE(max77857_ramp_table[0]),
> >> > + .ramp_reg = MAX77857_REG_CONT3,
> >> > + .ramp_mask = GENMASK(1, 0),
> >> > + .ramp_delay = max77857_ramp_table[0][0],
> >>
> >> This breaks the build with GCC 5.x through 7.x:
> >>
> >> drivers/regulator/max77857-regulator.c:312:16: error: initializer element is not
> >constant
> >> .ramp_delay = max77857_ramp_table[0][0],
> >> ^~~~~~~~~~~~~~~~~~~
> >> drivers/regulator/max77857-regulator.c:312:16: note: (near initialization for
> >'max77857_regulator_desc.ramp_delay')
> >>
> >> and clang:
> >>
> >> drivers/regulator/max77857-regulator.c:312:16: error: initializer element is not a
> >compile-time constant
> >> 312 | .ramp_delay = max77857_ramp_table[0][0],
> >> | ^~~~~~~~~~~~~~~~~~~~~~~~~
> >> 1 error generated.
> >>
> >> This relies on a GCC 8.x+ change that accepts more things as
> >> compile-time constants, which is being worked on in clang
> >>
> >(https://urldefense.com/v3/__https://reviews.llvm.org/D76096__;!!A3Ni8CS0y2Y!7B
> >eWxuzHgLzOprQA_madbvdR7hd0ZgmS73lUlDbgoxWUFWdDSIRXLnhyqLeRhu3uTaqpS
> >kzZKwc5pHA$ ). Since the kernel supports older
> >> compilers, this will have to be worked around somehow. Perhaps a define
> >> that can be used in both places?
> >
> >Was there any update on this? I do not mind sending a patch for this
> >myself if I have some sort of guidance on how you would prefer for this
> >to be fixed, should you be too busy to look into it.
> >
> >Cheers,
> >Nathan
>
> Hi Nathan,
>
> I thought that I should fix this issue after merging main branch that's why I did not send patch.
That is an understandable position but no, this issue should be fixed
before this change makes its way to Linus, not after.
> I sent patch v3 so should I send new patch as v4?
No, you should checkout Mark's branch that contains your patch and send
a new patch on top of it just fixing this issue, like the other two
patches that have already touched this driver:
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator.git/log/?h=for-6.6
https://git.kernel.org/broonie/regulator/c/2920e08bef609c8b59f9996fd6852a7b97119d75
https://git.kernel.org/broonie/regulator/c/541e75954cadde0355ce7bebed5675625b2943a8
There are GCC 7.x and earlier toolchains at
https://kernel.org/pub/tools/crosstool/ and LLVM toolchains at
https://kernel.org/pub/tools/llvm/ should need to reproduce and verify
the fix.
Cheers,
Nathan
