Re: musl - and menchmarking libc

[Rich Felker <dalias at aerifal dot cx>]

On Thu, Sep 06, 2012 at 04:50:22PM +0000, Joseph S. Myers wrote:
> On Thu, 6 Sep 2012, Rich Felker wrote:
> 
> > One could nit-pick about whether the "a header" condition disallows
> > this practice due to off_t, but my understanding of the intended
> 
> I think it clearly disallows self-declaring functions whose types involve 
> off_t

I think it's not so clear. Your reading may be correct by a strict
reading of the wording, but the intent seems sufficiently different
that this would call for a request for interpretation.

> (but a few functions such as "open" have 64-bit versions despite not 
> using such types - and unlike declaring functions with the type off_t in 
> fact has on the given implementation, it does seem to be allowed to 
> declare open and call it with the value O_RDONLY in fact has on the given 
> implementation - which could be passed in from another translation unit 
> that does include the headers).

Indeed. So I think we've found at least one case where a conforming
program can break under glibc's current handling of 64-bit-off_t mode
due to the need for the associated headers to remap functions.

Rich