Greg,
I agree that %d, %u, %x format specifiers be used but only with 'int',
'unsigned int', or 'signed int' type arguments. The issue here is that
%x is insufficient for uint32_t arguments in code built with
arm-none-eabi-gcc since the compiler defines __UINT32_TYPE__ as long
unsigned int (resulting in uint32_t to be typedef'd as 'long unsigned
int') that requires 'l' in corresponding format specifier. I'm surprised
I don't see more format specifier warnings while building for stm32...
Outside of unsized unsigned/signed int arguments, all other argument
types to printf formatting functions should use their corresponding
format specifiers from <inttypes.h>. Given the number of calls this is
going to take a while and require multiple pull requests (there are 400+
files with '%08x' in them to analyze)...
On 2/21/26 19:47, Gregory Nutt wrote:
%u should always be used but with type unsigned int only. Type unsigned int
may be 16, 32, or 64 bits wide. For portability the types in inttype.h should
always be used with fixed width integer types.
________________________________
From: Matteo Golin <[email protected]>
Sent: Saturday, February 21, 2026 5:54 PM
To: [email protected] <[email protected]>
Subject: Re: Non-portable printf format specifiers
Hi Peter,
Honestly, I think it's probably best practice to use the formatting macros
everywhere. The reason it hasn't been noticed in some of the other
locations you mentioned is because those platforms are architecture
specific where the warning doesn't apply. I'm not sure exactly why it
differs since uint32 is a fixed width, but I've seen this in places that
use %lu for unsigned int instead of %u, etc.
I would go for modifying whatever you can find with the macros if you're
willing, it can't hurt.
Matteo
On Sat, Feb 21, 2026, 6:12 PM Peter Barada <[email protected]> wrote:
While adding button support to nucleo-h743zi2:jumbo, I ran in six of the
same warnings in drivers/usbhosty/usbhost_storage.c that are independent
of the changes I made; here's the first of six warnings in the file:
usbhost/usbhost_storage.c: In function 'usbhost_dumpcbw':
usbhost/usbhost_storage.c:482:9: warning: format '%x' expects argument of
type 'unsigned int', but argument 3 has type 'uint32_t' {aka 'long unsigned
int'} [-Wformat=]
482 | uinfo(" signature: %08x\n", usbhost_getle32(cbw->signature));
| ^~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| |
| uint32_t {aka long unsigned int}
usbhost/usbhost_storage.c:482:26: note: format string is defined here
Fixing the warnings are easy in this case - replace '"%08x\n"' with
'"%08" PRIx32' where the corresponding argument is a uint32_t.
Out of curiosity I greped for '%08x' in the tree the tree and see other
places where that format is used with uint32_t corresponding argument
(e.g arch/arm/src/sama5/sam_nand.h::nand_getreg,
arch/arm/src/am335x/am335x_can.c::can_printreg, etc).
Question is should I create a pull request that makes this substitution
in just drivers/usbhost/usbhost_storage.c or should it include more
substitutions of '"%08x"' with '"%08" PRIx32 "' more/all cases where
%08x format is used with uint32_t typed argument?
--
Peter Barada
[email protected]
--
Peter Barada
[email protected]
