Re: On linux too (was Re: Base compilation broken on NetBSD)

[Richard Frith-Macdonald]

On 2 Nov 2012, at 13:50, Wolfgang Lux wrote:

> Wolfgang Lux wrote:
> 
>> I was going about to write the test, but then I found that autoconf already 
>> has a predefined test for it (obviously, we are not the only ones struggling 
>> with this dichotomy :-)
>> 
>> -- Macro: AC_FUNC_STRERROR_R
>>    If `strerror_r' is available, define `HAVE_STRERROR_R', and if it
>>    is declared, define `HAVE_DECL_STRERROR_R'.  If it returns a `char
>>    *' message, define `STRERROR_R_CHAR_P'; otherwise it returns an
>>    `int' error number.  The Thread-Safe Functions option of Posix
>>    requires `strerror_r' to return `int', but many systems
>>    (including, for example, version 2.2.4 of the GNU C Library)
>>    return a `char *' value that is not necessarily equal to the
>>    buffer argument.
>> 
>> So instead of using
>> AC_CHECK_FUNCS(strerror_r strerror)
>> in configure.ac, we should be using
>> AC_CHECK_FUNCS(strerror)
>> AC_FUNC_STRERROR_R
>> and add STRERROR_R_CHAR_P to config.h.in and use it to select a suitable 
>> variant of the code in the _systemError: method.
> 
> I have implemented this. The configure script checks for the variant of 
> strerror_r and the _systemError: method now contains code for the glibc 
> variant of strerror_r. Doing that, I had to remove the code from common.h 
> that was undefining _GNU_SOURCE and conditionally defining _XOPEN_SOURCE, as 
> that could invalidate the result of the autoconf test. I tested this on both 
> OS X (using the POSIX variant) and Linux (using the glibc version) and this 
> seems to work. I think it should also work for systems where strerror_r is 
> not defined at all and we define our own, POSIX compatible version, but I 
> have no system to test this.

Thnaks Wolfgang, that looks good.