Regression: tests failing on Solaris 10 for >= 2.0.23

[Laurent Blume]

Le 2014/08/14 20:01 +0200, Werner Koch a écrit:
> Hi!
>
> It might be that you are running an older version of gpg-agent which
> does not implement AGENT_ID and in this case the debug code might be
> triggered with a NULL for the text value.

I don't quite understand that part: there is no gpg-agent running when I 
start the tests, and I expect them to test the newly built one, not the 
currently installed one?

> Can you please try the patch for libassuan below.

I tried and it fixes the issue, all tests now succeed.

> However, I doubt that this is the reason for Solaris because I assume
> that printf ("%s", NULL) does not bail out on there but I am not sure.
> The AIX man page explictly remarks that it is undefined (as stated by
> Posix).  Time to move our own printf code from GnuPG to libgpg-error and
> have libassuan et al use that code too.  And we have the guarantee that
> "%zu" and such work.

Solaris tends to be picky about such things by default. It can be 
overruled to provide a Linux-like behaviour, but it's considered to be 
for debugging-only:

«The user compatibility library /usr/lib/0 at 0.so.1 provides a mechanism 
that establishes a value of 0 at location 0. Some applications exist 
that erroneously assume a null character pointer should be treated the 
same as a pointer to a null string. A segmentation violation occurs in 
these applications when a null character pointer is accessed.»

http://docs.oracle.com/cd/E19253-01/816-5165/ld.so.1-1/index.html

Thanks for your help! I'll push a patched libassuan for now.

Laurent