Re: [Fink-devel] libapr.0-dev-1.4.5-1 hardcodes fink's sed in itslibtool scripts:

[Daniel J. <dan...@gm...>]

On Jul 25, 2011, at 4:09 PM, Daniel Macks wrote:

> On Mon, 25 Jul 2011 15:04:51 -0400, Alexander Hansen  wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>> 
>> It doesn't actually carry any dependency on sed itself, but when I tried
>> to build a package that depends on libapr.0 (libserf0-0.7.0-3
>> ), I got the following error:
>> 
>> make LDFLAGS="-L/sw64/lib/system-openssl/lib -L/sw64/lib"
>> /sw64/share/apache2/build-1/libtool --silent --tag=CC --mode=compile gcc
>> - -g -O2  -DDARWIN -DSIGPROCMASK_SETS_THREAD_MASK -no-cpp-precomp
>> - -DDARWIN_10 -I/sw64/lib/system-openssl/include  -I.
>> - -I/sw64/include/apache2   -I/sw64/include/apache2 -I/sw64/include  -c -o
>> buckets/aggregate_buckets.lo buckets/aggregate_buckets.c && touch
>> buckets/aggregate_buckets.lo
>> /sw64/share/apache2/build-1/libtool: line 390: /sw64/bin/sed: No such
>> file or directory
> 
> [etc]
> 
>> 
> 
> > Right at the beginning of /sw64/share/apache2/build-1/libtool, after the
>> GPL statement, I have:
>> 
>> # A sed program that does not truncate output.
>> SED="/sw64/bin/sed"
>> 
>> # Sed that helps us avoid accidentally triggering echo(1) options like -n.
>> Xsed="/sw64/bin/sed -e 1s/^X//"
> 
> The libtool14 package had a similar problem, which was fixed by passing
> 
>   lt_cv_path_SED="/usr/bin/sed"
> 
> to ./configure in CompileScript. Maybe that works here too?

It does and I've now fixed it. It'd be nice if configure didn't use two different means of finding sed. The first one is overridden by setting SED but the second one ignores that and requires lt_cv_path_SED to be set. :p

Daniel