[BusyBox] Why glibc doesn't like busybox sed.

Erik Andersen andersen at codepoet.org
Mon Jan 26 07:05:37 UTC 2004


On Sun Jan 25, 2004 at 12:02:39PM -0600, Rob Landley wrote:
> So here's the snippet in the ./configure file for glibc-2.3.2 that's rejecting 
> busybox's sed (modulo cut and past wordwrap/whitespace damage):
> 
>   # Found it, now check the version.
>   echo "$as_me:$LINENO: checking version of $SED" >&5
> echo $ECHO_N "checking version of $SED... $ECHO_C" >&6
>   ac_prog_version=`$SED --version 2>&1 | sed -n 's/^.*GNU sed version 
> \([0-9]*\.[0-9.]*\).*$/\1/p'`
>   case $ac_prog_version in
>     '') ac_prog_version="v. ?.??, bad"; ac_verc_fail=yes;;
>     3.0[2-9]*|3.[1-9]*|[4-9]*)
>        ac_prog_version="$ac_prog_version, ok"; ac_verc_fail=no;;
>     *) ac_prog_version="$ac_prog_version, bad"; ac_verc_fail=yes;;
> 
> Notice that in order for us to fake out this test, we would have to reply with 
> the string "GNU sed version 3.1".  I.E. we have to claim to not only be a 
> Sed, but to specifically be _GNU_ sed.  A fully compatible sed implementation 
> isn't sufficient, they test for their brand of cola.
> 
> I think the correct fix here is to shoot this test in the head with a bazooka 
> full of broken glass and rusty nails.  Opinions?

<reading the glibc configure script>

This certainly looks like a case of someone being exceptionally
lazy.  The test in question is completely wrong-headeds.  Why
can't people build on i.e. freebsd or solaris with their native
sed versions?  If there is some specific sed feature they are
looking for, that specific feature should be tested for.

So yes, I agree with you fully on the merits of this particular
test,

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--



More information about the busybox mailing list