[darcs-users] the problems with autoconf

[zooko]

On May 15, 2008, at 9:41 AM, David Roundy wrote:

> On Thu, May 15, 2008 at 6:54 AM, zooko <zooko at zooko.com> wrote:
>> 1.  The current darcs autoconf scripts are not reliable at figuring
>> out, at configure time, what is going to be available at compile
>> time.  For example, ./configure will tell me that libcurl is
>> available, with various features, and then at compile time I will get
>> compile time errors saying that libcurl is not available.
>
> Did you try using the macro we created that handled your libz problem?
> If so, perhaps a patch and a bug report would be in order?

I did (and I spent a number of hours on it), but that macro is for  
compiling Haskell code which uses C code through FFI, where darcs's  
use of libcurl is pure C code (hscurl.c) using a libcurl.a, so I  
tried to make a variant of your macro which tested a C program  
instead of a Haskell program, but that macro that I made caused  
autoconf to go into an infinite loop and use up all of my RAM.  I  
couldn't debug my C-code-only macro before I gave up in despair.

Pekka Pessi has mailed me a patch that creates such a macro, so I  
will try it out next time I have time to work on this.

But note that currenty the configure script doesn't get anywhere  
close to testing anything, because it stops immediately when run  
saying that the C compiler cannot create executables.  (Because there  
is no C compiler on that machine, because the presence of a C  
compiler would only give a false sense of success to autoconf, since  
such a C compiler wouldn't actually be used to compile darcs.)

I couldn't figure out which instruction in darcs's configure.ac  
causes the resulting configure to do this test of a C compiler,  
before giving up in despair.

Thanks,

Zooko