On Fri, Sep 5, 2025 at 12:00 PM Peter Foley <pefo...@google.com> wrote:
>
>
> On Fri, Sep 5, 2025 at 4:48 AM Richard Henderson <
> richard.hender...@linaro.org> wrote:
>
>> On 9/5/25 09:31, Paolo Bonzini wrote:
>> > On 9/4/25 17:21, Peter Foley wrote:
>> >> We're getting errors about this:
>> >> linux-user/elfload.c:2770:36: error: use of undeclared identifier
>> 'MAP_FIXED_NOREPLACE'
>> >
>> > Queued, thanks.
>> >
>> > Paolo
>> >
>> >>
>> >> Signed-off-by: Peter Foley <pefo...@google.com>
>> >> ---
>> >> linux-user/elfload.c | 1 +
>> >> linux-user/mmap.c | 1 +
>> >> linux-user/syscall.c | 1 +
>> >> 3 files changed, 3 insertions(+)
>> >>
>> >> diff --git a/linux-user/elfload.c b/linux-user/elfload.c
>> >> index
>> 26c090c95d3e90ad4a23a927267e4106f68975b0..edbacf041f25d88472c95efb4eb9bdccd81e9902
>>
>> >> 100644
>> >> --- a/linux-user/elfload.c
>> >> +++ b/linux-user/elfload.c
>> >> @@ -5,6 +5,7 @@
>> >> #include <sys/prctl.h>
>> >> #include <sys/resource.h>
>> >> #include <sys/shm.h>
>> >> +#include <linux/mman.h>
>>
>> There's no reason to include <linux/mman.h>.
>>
>> This value is provided by <sys/mman.h> directly for musl and by
>> <sys/mman.h> via
>> <bits/mman-map-flags*> for glibc.
>>
>> Are you using some other libc? You're solidly in non-standard territory
>> there.
>> Perhaps that other libc needs updating.
>>
>
> We're using glibc 2.27.
> I definitely saw build failures earlier without this patch, but I'm now
> unable to reproduce them at head.
> So maybe there was an issue with our headers at an earlier point that has
> since been fixed?
> I'll revert this patch from our local fork and you should probably drop it
> upstream as well.
>
>
Oh, I figured out why I couldn't reproduce the issue.
We had a separate patch we were carrying that added linux/mman.h
to linux-user/user-mmap.h.
If I revert both patches, I see this:
third_party/qemu/linux-user/mmap.c:602:40: error: use of undeclared
identifier 'MAP_FIXED_NOREPLACE'
602 | if (start || (flags & (MAP_FIXED | MAP_FIXED_NOREPLACE))) {
>
>>
>> r~
>>
>
