Re: [bug-recutils] [Platform-testers] recutils pre-release 1.4.91 in alpha.gnu.org

[Bruno Haible]

Hi José,

>     * Linux/PowerPC (both 32-bit and 64-bit)
>     * Linux/SPARC (64-bit)
>     * MacOS X 10.5 (both 32-bit and 64-bit)
>     * NetBSD 5.1
>     * OpenBSD 4.9
>     * Cygwin 1.7.9
>     
>       recsel-confidential fail (see recsel-confidential.diff)
>       recsel-confidential-fex fail (see recsel-confidential-fex.diff)
>       recsel-confidential-fex-value fail (see 
> recsel-confidential-fex-value.diff)
>       recsel-confidential-num fail (see recsel-confidential-num.diff)
>     FAIL: recsel.sh
>     
>     Find the files attached.
> 
> It looks like in those systems libgcrypt is not available, so the output
> of the tests contains the encrypted strings.  I must put in place
> something in the tests to recognize such cases to not run the encryption
> related stuff.

Yes, skipping these parts of the tests sounds like the right fix.

> I imported the strcasestr module from gnulib.  The recfiles are supposed
> to contain UTF-8 strings.

OK. On UTF-8 strings strstr() and strcasestr() are reliable.

> Those ones are due because in those systems %zd does not work in printf
> strings.  I am using %zd in order to avoid a warning in recent versions
> of gcc regarding printf and size_t.  Do you know about a portable and
> safe format string to print size_t values?

Sure. Use the module 'printf-posix' or 'fprintf-posix' or 'sprintf-posix'
(whichever is appropriate) from gnulib, and the problem will be solved.

Alternatively - if you find these *printf modules from gnulib too heavy -,
you can also cast the argument from size_t to 'unsigned long' and use %lu
instead of %zu. Or, from ssize_t to 'long', and use %ld instead of %zd.
Then you don't need the *printf-posix modules from gnulib.

Bruno