This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Coreutils head and tail problems

On Tue, Aug 26, 2003 at 09:40:20AM -0700, Zack Weinberg wrote:
> Joe Buck <> writes:
> > On Tue, Aug 26, 2003 at 08:58:13AM -0700, Zack Weinberg wrote:
> >> These are deliberate and will not change.  GCC has to maintain
> >> backward compatibility with platforms which do not support head -n1,
> >> tail -n1, etc ... the issues you have found are merely the tip of the
> >> iceberg.
> >
> > I agree that the fact that the coreutils people broke gcc builds was a
> > mistake; even if we decide to change, we can't retroactively fix gcc
> > builds that we have already released.  But which systems don't support
> > head -n1 and tail -n1?  I've scouted about 8 OS flavors I have access to
> > (Solaris, HP-UX, SGI-MIPS, Alpha OSF, FreeBSD, Linux, AIX) and they all
> > handle it.  SunOS4, perhaps?
> My recollection was that at least some still-supported versions of
> HP-UX did not support this.  I'll drop my objection to change if
> someone audits every supported host architecture (and "X, Y, Z don't
> support it but they were EOLed in 1992 so why are we still supporting
> them anyway?" would be an acceptable outcome in my book).

Another issue is that if Posix has deprecated "head -1", eventually other
tools providers will stop supporting it, so autoconf will have to adapt.
It might turn out that the only portable approach will be to do a
configure-time test.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]