[Pvfs2-developers] state machine compiler tweaks

[Sam Lang]

On May 30, 2006, at 9:21 AM, Pete Wyckoff wrote:

> walt at clemson.edu wrote on Tue, 30 May 2006 09:28 -0400:
>> Which is more common in compilers:
>>
>> static struct foo foobar = {
>> 	field1: value1,
>> 	field4: value4
>> 	};
>>
>> OR what's shown below:
>>
>> static struct foo foobar = {
>> 	.field1 = value1,
>> 	.field4 = value4
>> 	};
>>
>> The first is used by the kernel code, and  has been around for  
>> quite a
>> while.  I'm not sure what the second form gives us that the first one
>> doesn't.  Unless there is a good reason otherwise we should probably
>> stick with the most widely supported.
>
> The first is a gcc-only extension, obsolete since gcc 2.5 according
> to their manual.  The latter is part of the ISO C99 spec.  If we
> were going to convert from the unnamed version to anything, I'd have
> a strong preference for going with the spec.
>
>> Murali Vilayannur wrote:
>>> Oops that was my bad..I did not realize that this syntax is not  
>>> supported
>>> by many compilers! I was once bitten by initializing a struct
>>> incorrectly, and I committed that change since I assumed that a  
>>> user would
>>> not be able to compile the kmod without having a compiler that  
>>> understands
>>> that syntax.
>
> The kernel code is inherently tied to gcc, so designated
> initializers causes no problem.  But the few places in dist-*.c that
> went in last month might affect non-linux non-gcc clients or servers.
> Looking back at old gcc versions, even ancient 2.5 supported this
> feature.  Scanning web pages for sun and ibm show that they have had
> support for some time too.
>
> I think our only problem would be with an old vendor-supplied
> compiler.  And my feel is that we should not bother supporting such
> if it reduces our code quality.  For the next release, let's leave
> in the few places (dist-*.c, and soon statecomp) where these are
> used and see if anybody complains.  It will be easy enough to back
> it out.  Other opinions?
>

I remember seeing other gcc specific features used in the pvfs2 code,  
such as varargs for macros, so I had thought that they were requiring  
gcc anyway.  Does intel support varargs for macros?

> 		-- Pete
> _______________________________________________
> Pvfs2-developers mailing list
> Pvfs2-developers at beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
>