libhttp

[Alan DuBoff]

Ian Zimmerman wrote:

> Now I guess it's time to explain why I hate automake :-) I don't
> necessarily claim that the following reasons apply to Alan or his
> project.  YMMV.
> ...
> 1 - Automake actually considers _too many_ oddball cases
> 2 - The generated makefile includes a dependency of Makefile.in on
>     configure.in.
> 3 - It is way too C-centric.
> 4 - It assumes a recursive make strategy.

Ian,

As I mentioned previously, libhttp was the first piece of code I've personally
used automake/autoconf/libtool with, so I'm certainly no expert. And it's
kinda messy as it's all in one directory.

Even with the items you list above, it still seems to be the best solution,
whether these are a real problem or not, that I'm not certain. Being too
c-centric for instance, well, duh! UNIX/Linux is very c-centric at the command
line, it doesn't seem odd that some of the tools would be...

I will say that automake/autoconf/libtool is not the most straight forward to
understand, and one thing bothers me with them, in that you can't test a
specific platform without taking the resulting code to the platform. IOW, even
if I want to make sure my code runs on some flavor of UNIX, such as Solaris or
HP-UX, it still requires the code be taken there and run to see if it will
configure and build. Some SVR4 flavors of UNIX don't include the socket or nsl
libs as a part of the standard libraries, but you wouldn't know that until the
code is configured and built on that platform, at which time it blows
up...with embedded development for instance, 3 platforms are typically used
for the host development platform, Linux x86, Linux PPC, and Solaris SPARC.
It's hard to know for certain that the resulting distribution will configure
and build on each platform without taking it and actually testing it on each
one of them.

-- 

Alan DuBoff
Software Orchestration, Inc.