-p considered problematic (was Re: how to proceed past errors?)
jeremy at lavergne.gotdns.org
Tue Jul 31 08:05:46 PDT 2012
> Why is this considered problematic? My understanding is that this
> won't allow a port to be installed if it's dependencies fail, but will
> continue to build as many requested ports as is possible.
> ie: port-a depends on port-b, port-c, and port-d
>> port install port-a
> would install port -b, port-c fails, and port stops
>> port -p install port-a
> would install port-b, port-c fails, install port-d, doesn't install
> port-a because deps aren't satisfied.
> I've been using this flag in an automated process for quite some time.
> What problems should I be aware of with using this switch?
The only issue that comes to mind is missing/incorrect dependencies. A package may install (from binary or source) but there may be issues because it built or rebuilt differently based on availability of an opportunist dependency.
We've seen some maintainers write library dependencies as build dependencies, which of course would be missing when installing in the default, binary mode. There's also the case where a dependency was already installed on the developer's machine and the software would link opportunistically with it: this results in no error for anyone but may result in missing functionality.
I don't think that's a problem with -p, however it will be more likely to manifest itself that way; it's more of a bug with the port that's allowing that opportunist linking or otherwise incorrectly declared dependencies.
More information about the macports-users