Re: CVS automake testsuite failures on NetBSD i386 2.0F

[Nicolas Joly]

On Sun, Jun 06, 2004 at 10:40:55PM +0200, Alexandre Duret-Lutz wrote:
> >>> "Nicolas" == Nicolas Joly <address@hidden> writes:
> 
>  Nicolas> and noticed 2 failures :
> 
>  Nicolas> FAIL: tar.test
>  Nicolas> FAIL: tar2.test
> 
>  Nicolas> According to the attached traces, it seems that tar archives program
>  Nicolas> detection does not work very well, leading in an empty `am__tar'
>  Nicolas> variable :
> 
> [...]
>  Nicolas> checking how to create a ustar tar archive... gnutar plaintar pax 
> cpio none
> [...]
> 
> I can see this on NetBSD 1.6.1 too:
> 
> $ for i in a b c; do echo $i; done
> a
> b
> c
> $ for i in ${x-a b c}; do echo $i; done
> a b c
> $ x='x y'
> $ for i in ${x-a b c}; do echo $i; done
> x
> y
> 
> This looks like a bug in NetBSD's sh to me.  POSIX says field
> splitting should occur after parameter expansion, and replacing
> ${x-a b c} by "x y" or "a b c" is part of parameter expansion.
> So where the value comes from should not make any difference.

With Tru64 unix and Solaris `/bin/sh', things look worse :

$ for i in a b c; do echo $i; done
a
b
c
$ for i in ${x-a b c}; do echo $i; done
bad substitution
$ x='x y'
$ for i in ${x-a b c}; do echo $i; done
bad substitution

even zsh does not seems to behave very well :

address@hidden [~]> zsh --version
zsh 4.2.0 (i386-unknown-netbsdelf2.0F)
address@hidden [~]> for i in a b c; do echo $i; done
a
b
c
address@hidden [~]> for i in ${x-a b c}; do echo $i; done
a b c
address@hidden [~]> x='x y'
address@hidden [~]> for i in ${x-a b c}; do echo $i; done
x y

> May I let you report this to the NetBSD people?  I'm not a
> NetBSD user myself, while it looks like you're running bleeding
> edge.

Will do.

> I'm installing this following patch for Automake.

Now all tests are successful :
=======================================================
All 465 tests behaved as expected (3 expected failures)
(81 tests were not run)
=======================================================

Thanks a lot.

-- 
Nicolas Joly

Biological Software and Databanks.
Institut Pasteur, Paris.