Re: Build error in 6.16-rc1 in drivers/mfd/88pm860x-core.c

[Mario Limonciello]

On 6/8/25 8:58 PM, Peter Schneider wrote:
> Hi all,
>
> I just tried to test-build 6.16-rc1 on my good old 2-socket Ivy Bridge 
> Xeon E5-2697 v2 server, running Proxmox/Debian 12 (Bookworm) with GCC 
> 12.2. I received the following build error:
>
>    CC      drivers/mfd/88pm860x-core.o
> drivers/mfd/88pm860x-core.c: In function 'device_irq_init':
> drivers/mfd/88pm860x-core.c:576:29: error: unused variable 'node' [- 
> Werror=unused-variable]
>    576 |         struct device_node *node = i2c->dev.of_node;
>        |                             ^~~~
> cc1: all warnings being treated as errors
> make[4]: *** [scripts/Makefile.build:287: drivers/mfd/88pm860x-core.o] 
> Error 1
> make[3]: *** [scripts/Makefile.build:554: drivers/mfd] Error 2
> make[2]: *** [scripts/Makefile.build:554: drivers] Error 2
> make[1]: *** [/usr/src/linux/Makefile:2006: .] Error 2
> make: *** [Makefile:248: __sub-make] Error 2
>
>
> My config file is attached. I always build with CONFIG_WERROR=y (same as 
> I do in all of my regular test builds of Greg's stable-RC releases).

I do too; I'm pretty surprised that that the build robots didn't catch this.

> I'm not a Kernel developer nor a C programmer, I only have superficial 
> knowledge of C (but I love to test and to tinker), so I don't quite 
> understand this warning/error message. Becauses some lines below, the 
> node variable is used as a paramter to the of_fwnode_handle macro (from 
> include/linux/of.h):
>
>          irq_domain_create_legacy(of_fwnode_handle(node), nr_irqs, chip- 
>  >irq_base, 0,
>                                   &pm860x_irq_domain_ops, chip);
>
> Does that not count as variable usage to the compiler? 
It's not because there is a macro when CONFIG_OF is not defined that 
assigns it as NULL.  So to the compiler there is no variable use.

> Digging in 
> somewhat deeper, I found that probably the patch from [1] is missing in 
> 6.16 and should have been included?!

It was sent during the merge window, it's not a surprise that it wasn't 
picked up.  I just sent it in case they had a chance to grab it because 
not being able to compile with CONFIG_WERROR means other things can more 
easily get missed too.

> When I apply that patch, my build succeeds.
>
> [1]: https://lore.kernel.org/all/20250602201008.1850418-1- 
> superm1@xxxxxxxxxx/
>
>
> Beste Grüße,
> Peter Schneider