Re: eCAP configure tests

On Tue, 2008-10-14 at 22:06 +1300, Amos Jeffries wrote:

> On the eCAP side. I've found an issue with its configure tests. Seems to
> me that the MSG_ERROR and MSG_FATAL are the wrong way around.

I did not find MSG_FATAL in autoconf documentation. The code is using
AC_MSG_ERROR and AC_MSG_FAILURE, both of which are "fatal" to the
configure process.

> Should be:
> * FATAL if ecap listed with explicit disable-loadable

The code is using AC_MSG_ERROR, which is fatal. I think it is the right
macro to use when explicit configure options contradict each other. Do
you agree? If not, what macro would you use?

> * ERROR with failover to disable, if ecap listed but libraries missing.

The code is using AC_MSG_FAILURE, which is an AC_MSG_ERROR wrapper. Are
you suggesting that we use AC_MSG_WARN and implicitly disable eCAP
support that the user has explicitly enabled?

> Without a sane failover we can't do portable permutation testing of the
> ecap code asent libecap. The existing testbed does not handle fatal exit
> cases yet.

What is "portable permutation testing"?

Would Kinkie's suggestion to implicitly enable eCAP if there is libecap
installed solve the problem? I may prefer that because it does not
override explicit user configuration and because many, if not most,
users pay little attention to ./configure warnings.

IMO, if a user explicitly requested feature Foo and Foo cannot be
supported, we should fail rather than ignore the user request.

Thank you,

Alex.
Received on Tue Oct 14 2008 - 14:24:23 MDT