Re: failed tests AC_CHECK_LIB

[Keith Marshall]

On Sunday 21 September 2008 21:12:22 Matěj Týč wrote:
> Thank you, Brian, I have sent your message to the macro author, so
> now it is really easy for him to fix it for others as well.

Why don't you just adapt it yourself, to suit your own requirements?  
That after all, is the underpinning of the open source ethos -- the 
ability of any user to adapt code when necessary.

> Anyway, don't you think that the usability of AC_CHECK_LIB in the
> current form (not allowing inclusion of headers nor the precise
> specification of the function call) is quite low because of this
> calling convention issue?

This "calling convention issue" is specific to MS-Windows, which is a 
*very* minor target platform for autoconf; indeed native MS-Windows 
doesn't even provide a shell which is capable of running an autoconf 
configure script.

AC_CHECK_LIB is perfectly usable on the *nix platforms, where it is 
predominantly deployed.  It doesn't need to be made more complex, to 
suit a quirk of a minority usage case.

> Don't you, autoconf guys, plan something more complex, but more
> robust and allowing checking for methods in C++?

It isn't a C vs. C++ issue; it is an issue caused by a calling 
convention which is specific to MS-Windows.  It is also an issue 
which would be very difficult, if not impossible to work around in a 
generic manner -- as AC_CHECK_LIB does for __cdecl functions -- 
because not only is it necessary to include the appropriate function 
prototype, but then the test must call the function with an argument 
list which is customised to match that prototype.

Regards,
Keith.