Re: libtool max_cmd_len bits

[Mike Stump <mrs at apple dot com>]

On May 7, 2005, at 4:24 PM, Joseph S. Myers wrote:
> On Thu, 7 Apr 2005, Mike Stump wrote:
> > This pulls some bits from libtool to update  
> > lt_cv_sys_max_cmd_len...  this
> > improves -m64 multilib builds of libjava on darwin8 by enabling it  
> > to build in
> > one step.
>
> > +if test -n $lt_cv_sys_max_cmd_len ; then
>
> This use of "test" without quoted argument ($lt_cv_sys_max_cmd_len  
> may be
> empty at this point) causes problems configuring with those shells  
> which
> exit after "test: argument expected".  Those shells are buggy but the
> configure script still shouldn't be invoking test with just "-n" as  
> its
> argument; $lt_cv_sys_max_cmd_len should be quoted.

:-(  And I tested that case on my system, and it worked because my  
shell isn't `buggy'.

When more of libtool is imported, we won't need to do this.  I ran  
into problems trying to import a larger portion of libtool, in  
particular, the case *) bit above this code in libtool, and leaving  
that out, is essentially I think what breaks this when compared to  
raw libtool.

I checked in the below patch to fix this, as my intention wasn't to  
break configuration on such systems.  Thanks for your help in  
spotting this.

If someone wants to try their hand importing more of libtool, that  
would be great.

Attachment: libjava-7.diffs.txt
Description: Text document