On 1/7/2013 4:45 AM, Daniel Stenberg wrote:
> On Sun, 6 Jan 2013, John E. Malmberg wrote:
>
>> The configure test for AIX is producing a false positive.
>>
>> This is resulting in the _ALL_SOURCE macro being defined.
>>
>> As near as I can tell, I do not see any bad effects on VMS.
>>
>> It looks like the problem is that the test should produce a compile
>> failure to indicate AIX, and instead a compile success is being used.
>
> Not quite. The code for CURL_CHECK_AIX_ALL_SOURCE in acinclude.m4 does
> this:
>
> AC_EGREP_CPP([yes_this_is_aix],[
> #ifdef _AIX
> yes_this_is_aix
> #endif
> ],[
> AC_MSG_RESULT([yes])
> AC_DEFINE(_ALL_SOURCE)
> ],[
> AC_MSG_RESULT([no])
> ])
>
> Which means it'll preprocess that little snippet of code and check for
> 'yes_this_is_aix' in the output. That should only happen if _AIX is
> defined when the preprocessor runs... Or can you think of any other
> explanation? Can you figure out why it detects that in the output for
> your test?

Yes, because I have full listings enabled in order to debug the
configure and build process. Which means that the compiler lists out
even the lines that are not compiled.

Regards,
-John
wb8tyw_at_qsl.net

-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html
Received on 2013-01-07