Re: [RFC PATCH] Add -Werror flag when building GNU/HURD.

On 7/28/26 6:38 AM, Bradley Morgan wrote:
> On 28 July 2026 11:31:23 BST, Mikhail Karpov <[email protected]>
> wrote:
>> Hello,
>>
>> When compiling GNU/Hurd, quite a few warnings were generated. I fixed some
>> of
>> them, and now it compiles with flags:
>> CFLAGS += -std=gnu99 $(gnu89-inline-CFLAGS) -Wall -Werror \
>>                      -Werror=old-style-definition \
>>                      -Werror=implicit-function-declaration \
>>                      -Werror=strict-prototypes \
>>                      -g -O3 -fno-strict-aliasing $($*-CFLAGS)
>> Completes successfully.
>>
>> I have a few questions:
>> 1. Is this flag even necessary?
>> 2. Should I remove the -fno-strict-aliasing flag? I didn't remove it, so
>> the
>>     -Wstrict-aliasing warnings are ignored.
>> 3. Should I add the -Wextra flag?
>>
>> I agree that warnings are sometimes annoying, but I believe that catching
>> a
>> real error at compile time can save a lot of time searching for it in the
>> code. However, I'd still like to hear people's opinions on this matter.
>>
>> Now on to the patches.
>> I'm not sure about some of them, and they probably require more time to
>> study. For example, in pfinet/linux-src/include/linux/in_route.h there was
>> #define RTCF_NAT (RTCF_DNAT|RTCF_SNAT) (0x08000000|0x00800000), but
>> previously in tools-i686/include/net/route.h there was also #define
>> RTCF_NAT
>> 0x00800000. The same problem occurs with #define NET_MAX_FILTER 512 from
>> libbpf/bpf_impl.h and #define NET_MAX_FILTER 128 from
>> tools-i686/include/device/net_status.h.
>>
>> I simply removed unused variables and moved unused functions to #if 0.
>>
>> pfinet encountered a false-positive warning -Werror=array-bounds in the
>> queue
>> handling functions. I couldn't find a simple solution, so I simply added
>> -Wno-array-bounds to the pfinet Makefile.
>>
>> And a few places that were unclear to me:
>> The -Werror=maybe-uninitialized warning was in
>> pfinet/linux-src/net/ipv4/igmp.c because im->loaded = 0 was only set
>> within
>> CONFIG_IP_MULTICAST, and the igmp_group_added function checked im->loaded
>> for 0 when it was called.
>> The -Werror=address warning was in pfinet/linux-src/net/ipv4/devinet.c
>> because ifa->ifa_label is an array, not a pointer, so the if
>> (ifa->ifa_label)
>> condition is always true. I removed this check and left only
>> strcpy(ifr.ifr_name, ifa->ifa_label).
>> The Werror=shift-overflow warning was in the libps/ps.h file: #define
>> PSTAT_USER_BASE 0x20000000, and #define W_PSTAT_LOGIN (PSTAT_USER_BASE <<
>> 3),
>> so the result didn't fit into the int type. I added the ULL suffix to
>> 0x20000000 and that solved the problem.
>>
>> The fixing-structure-declaration-inside-parameter-list.patch patch applies
>> to
>> the libtirpc-1.3.6/tirpc/rpc/rpc_msg.h file. It must be applied
>> separately.
>>
>> Thanks,
>>
>>
>
> Hi, Where's the patch?
>
> Adding -WError is relatively a non issue, (as the user could build with
> W1,or W128 if they are feeling crazy) Forcing the user to build with
> warnings as errors makes the build(er) responsible for fixing
> someone elses BS

So I am not a Hurd developer, but I think this is a good idea to turn on 
WError be default.  Didn't Linus just turn on WError by default for the 
kernel?  I could be wrong.

Also, GCC 14 is WAY more strict about how you compile C (1).  I feel 
like the GCC developers are making C more strict to combat the 
competition from rust?

1)  https://gcc.gnu.org/gcc-14/porting_to.html   I learned about that 
url via reading guix's blog post: 
https://guix.gnu.org/en/blog/2026/the-64-bit-hurd/


Joshua

>
> If I see the patch, I'll have a look 😀
>
> Thanks!
>