httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ralf S. Engelschall" <>
Subject Re: cvs commit: apache-1.3 STATUS
Date Mon, 17 Aug 1998 06:22:29 GMT

In article <> you wrote:
> On Sun, 16 Aug 1998, Jim Jagielski wrote:

>> Ralf S. Engelschall wrote:
>> > 
>> > Sounds reasonable as a first step of improvement. The CFLAGS->#define are
>> > harmless because they are surrounded with #ifndef..#endif. But I'm not sure
>> > this solves the problem of having to #include some other headers for some
>> > headers we test. Or do you think they are then automatically included because
>> > of ap_config.h? 
>> Yep, that's what I'm thinking. Basically, if we include the "in-process"
>> version of ap_config_auto.h after ap_config.h then that's gotta be better.

> Yes, as long as you do not have any platforms with conflicting ordering
> requirements.  

> In theory you shouldn't have any such platforms, and they can be dealt
> with when they are found.

The other point is that usually when one tests for the _existence_ of a header
file one should not also test that its _working_ at the same time. In GNU
Autoconf for the headers tests only CPP (or CC -E) is used. No actual compile
is tried. This way the header dependencies can be ignored. I wanted to make it
this way inside TestCompile, too. But it cannot be done without adding a check
"how to run CPP stand-alone" to Configure.

But that seems the only correct solution to me: Configure has to not only find
the CC, it also has to find out how CPP can be called stand-alone. Then the
"TestCompile header" has to use CPP _only_! It just should test whether the
header _exists_. That's all.

                                       Ralf S. Engelschall

View raw message