eisentrp <[EMAIL PROTECTED]> writes:
> On 16 Jul 2000, Russ Allbery wrote:
>> My macro instead does *tzname = "UTC". I'm not aware of any platform
>> that uses these variables but doesn't allow the user to modify them, so
>> that seems to work.
> Seems reasonable, but the fact that you have to wonder in the first
> place makes this the seemingly less portable solution.
True. What about strlen(*tzname) if it really has to appear in a function
call? (Other autoconf tests, BTW, just use the variable without passing
it to any function, so I'm not sure even that's needed.)
> In particular I'd like to see a more generalized approach to "find out
> whether package X is available and how" (re OpenSSL etc.). Currently,
> every package rolls its own interface.
Yeah, that would be nice, although a few things like Perl and Python
require a lot of fairly specialized tricks.
I'm mostly going to be glad to clean up all the extraneous cruft in the
INN configure.in right now when we have things like tests for types that
allow inclusion of headers in the main autoconf release.
Oh, another annoyance -- the handling of --with-<blah> doesn't deal with
--without-<blah> in a way that's particularly easy. You can't just put
your code to enable <blah> in the "true" portion of AC_ARG_WITH, since if
someone explicitly uses --without-<blah>, that triggers the "true" portion
and just sets the argument to "no". This is rather counter-intuitive; I
wonder how many other packages have gotten this wrong at one point or
another.
--
Russ Allbery ([EMAIL PROTECTED]) <http://www.eyrie.org/~eagle/>
