RFE: new prefix for heimdal affects other ports

[MarcusCalhoun-Lopez (Marcus Calhoun-Lopez)]

In the recent revision, r35247, heimdal was installed into its own directory because "no other ports depend on it."
At least two ports inadvertently depends on it because it was found in the configure phase (as reported in #14697 and #12805).

If the reason for the change was to avoid conflict with the system kerberos, doesn't this conflict with MacPorts policy of using its own libraries?

Since I only use heimdal as part of other ports, my interest is more or less academic, but if, as indicated, it is "much nicer than the included MIT kerberos," shouldn't it be returned to the central location so that other ports can more easily use it?

[landonf (Landon Fuller)]

Heimdal and MIT Kerberos both place header files and binaries directly in the search path that conflict with each other, so I finally decided to create a separate prefix.

As a strong proponent for the "using our own libraries" policy, I find myself in the funny position of defending the exact opposite tack. On Mac OS X, Kerberos is a system service. It integrates with the Keychain, the GUI, etc. You can make Heimdal kerberos integrate with the system ticket store, but (I believe this is still the case), it doesn't do so by default, and so you wind up with dual ticket stores and libraries and configuration.

I could be convinced that I'm wrong, and that ports should depend on Heimdal kerberos, and we should just ensure that it works correctly with the system ticket store by default.

I am pretty sure about using a separate prefix though -- otherwise, too many things conflict in the same search space.

[MarcusCalhoun-Lopez (Marcus Calhoun-Lopez)]

Thank you for taking the time to explain.
For what it's worth, your reasons seem solid to me.

Personally, I do not have a strong motivation to get heimdal working with the system ticket store,
but I appreciate the insight into the kerberos setup.

[(none)]

Milestone Port Enhancements deleted