Re: libidn-1.30: Test failure on NetBSD

[Thomas Klausner]

Hi Simon!

On Wed, Jul 08, 2015 at 10:24:22PM +0200, Simon Josefsson wrote:
> Thomas Klausner <address@hidden> writes:
> 
> > Hi!
> >
> > lib/gltests/test-localename.c from libidn-1.30 fails to build on NetBSD:
> >
> >   CCLD     test-localename
> > test-localename.o: In function `test_locale_name_thread':
> > test-localename.c:(.text+0x23): undefined reference to `uselocale'
> > test-localename.c:(.text+0x73): undefined reference to `uselocale'
> > test-localename.c:(.text+0x109): undefined reference to `uselocale'
> > test-localename.c:(.text+0x191): undefined reference to `uselocale'
> > test-localename.c:(.text+0x1ce): undefined reference to `uselocale'
> > test-localename.o:test-localename.c:(.text+0x27c): more undefined 
> > references to `uselocale' follow
> > *** Error code 1
> >
> > The configure script already checks for this function, but the result
> > is ignored (from config.h):
> >
> > /* Define to 1 if you have the `uselocale' function. */
> > /* #undef HAVE_USELOCALE */
> >
> > Please do not reference uselocale() on systems where it doesn't exist.
> 
> Hello Thomas.  Thanks for the report.
> 
> I cannot reproduce this.  On my NetBSD 6.1.4 (a virtual machine
> downloaded from http://pretest.nongnu.org/), which doesn't have
> uselocale either, it builds fine.

Thanks for installing NetBSD and trying this yourself!
However, you'll need a newer version of NetBSD to reproduce this problem.

I'm running NetBSD 7.99.19/amd64, but newlocale() was added already in
2013, so you can also try it in a NetBSD 7 release candidate.

A snapshot of that is available at
http://nyftp.netbsd.org/pub/NetBSD-daily/netbsd-7/201507072130Z/amd64/installation/cdrom/
but they only last a couple of days, then you'll have to look in
http://nyftp.netbsd.org/pub/NetBSD-daily/netbsd-7/
for a newer one.

> The lib/gltests/test-localename.c code should use the HAVE_NEWLOCALE
> #define to check whether it should call uselocale or not, and on my
> NetBSD system config.h correctly have this:
> 
> /* #undef HAVE_NEWLOCALE */

I think that is exactly the problem -- it shouldn't check
HAVE_NEWLOCALE to decide if it uses uselocale(), but HAVE_USELOCALE
instead.

> You will have to debug a bit more, or give me some way of reproducing
> it.  Also, I'm currently in the process of releasing 1.31 so please try
> that version before debugging anything.

Do you need more information or is this enough?
If not, please let me know what else.

Thanks,
 Thomas