2.6.18-mm1: true/false enum in linux/stddef.h fails glibc-2.4 compile

2.6.18-mm1: true/false enum in linux/stddef.h fails glibc-2.4 compile

[Eric Buddington]

There is an enum contained in some recent -mm versions of
linux/stddef.h which seems to be horking my glibc-2.4 compile:

enum {
        false   = 0,
        true    = 1
};

One way or another (I can't find where), 'true' and 'false' are
getting defined to 1 and 0, turning the above into enum { 0=0, 1=1 },
which though undeniable is not compilable.

This enum shows up in 2.6.18-rc5-mm1 and 2.6.18-mm1, but not in
2.6.17-mm6.

Whether this is the fault of glibc, gcc, or the kernel is a topic I
wouldn't touch with a 3m pole, even if I had one. I've submitted the
issue to the glibc bugzilla as well:

http://sources.redhat.com/bugzilla/show_bug.cgi?id=3301

Any thoughts would be appreciated.

Thanks,

Eric

Re: 2.6.18-mm1: true/false enum in linux/stddef.h fails glibc-2.4 compile

[Arjan van de Ven]

On Wed, 2006-10-04 at 11:51 -0400, Eric Buddington wrote:
> There is an enum contained in some recent -mm versions of
> linux/stddef.h which seems to be horking my glibc-2.4 compile:
> 
> enum {
>         false   = 0,
>         true    = 1
> };
> 
> One way or another (I can't find where), 'true' and 'false' are
> getting defined to 1 and 0, turning the above into enum { 0=0, 1=1 },
> which though undeniable is not compilable.

I think you're making the mistake of using kernel headers for
userspace...... rather than the cleaned up headers.

Re: 2.6.18-mm1: true/false enum in linux/stddef.h fails glibc-2.4 compile

[Richard Knutsson]

Arjan van de Ven wrote:
> On Wed, 2006-10-04 at 11:51 -0400, Eric Buddington wrote:
>   
>> There is an enum contained in some recent -mm versions of
>> linux/stddef.h which seems to be horking my glibc-2.4 compile:
>>
>> enum {
>>         false   = 0,
>>         true    = 1
>> };
>>
>> One way or another (I can't find where), 'true' and 'false' are
>> getting defined to 1 and 0, turning the above into enum { 0=0, 1=1 },
>> which though undeniable is not compilable.
>>     
>
> I think you're making the mistake of using kernel headers for
> userspace...... rather than the cleaned up headers.
>   
I think so too. glibc-2.4 #includes stdbool.h which have false/true 
defined as 0/1.

Richard Knutsson