[Box Backup] bbackupquery strange issue

[Ben Summers]

On 10 May 2006, at 14:43, Chris Wilson wrote:

> Hi Ben,
>
> On Wed, 10 May 2006, Ben Summers wrote:
>
>>> Ben, perhaps if we detect HAVE_VALID_DIRENT_D_TYPE and compile  
>>> with it enabled, then we should add a check that d_type is  
>>> correct for the "." and ".." directories, and maybe for a known  
>>> file as well?
>>
>> Yes if it's going to break silently otherwise. Tracking Linux as a  
>> target is a nightmare.
>
> I guess you could say that "tracking *BSD is a nightmare" as well,  
> because the diversity between Linuxes and the diversity between the  
> BSDs (including Darwin) is probably similar :-)

Not quite a fair comparison. Targeting "Linux" is targeting an  
effectively unknown number of platforms. Targeting BSDs is known  
number of platforms. The thing about every installation of Linux  
being a surprise is the thing.

>
>>> Or else just use stat() on all platforms?
>>
>> Doesn't sound the nicest plan ever.
>
> What's wrong with stat()? d_type is not in the POSIX standard so we  
> shouldn't really use it for portable code (IMHO).

Efficiency. The point of configure is to allow us to use the good  
stuff where we can without having to drop to the lowest denominator.  
With this argument you'd have us all using select()...

Ben