This should not be the main difference between depends_lib and depends_run as I 
see it.  The main difference between those should be architecture.

If I depends_run on port:bzip2, I just care that the bzip2 executable be 
present, and I can run it.  I don't care what architectures it has.

If I depends_lib on port:bzip2, I intend to link against libbz2 and need to 
make sure that the installed architectures match the build architectures.

Thus ports only in depends_run should be in depends_skip_archcheck.  Whether 
that is implemented yet or not, I'm not certain, but that is the intention.

--Jeremy

> On Jan 17, 2015, at 06:47, Jeremy Lavergne <[email protected]> wrote:
> 
> Thanks for clarifying the difference, Clemens.
> 
>>> Looks like the bug is in portutil.tcl.  depends_run should be included
>>> as a dependency of build.  I'll just change it to a _lib dependency.
>> 
>> If we do that, that would remove the last difference between depends_lib
>> and depends_run. I was under the impression that this difference was why
>> we even had both depends_run and depends_lib. As you might know, we do not
>> currently depend on depends_lib only containing libraries (and IMO that
>> doesn't make sense because we could easily figure out dynamic library
>> dependencies using a couple of lines of code automatically, and static
>> library dependencies are usually not even listed in depends_lib, but
>> depends_build).
> 
> I vaguely remember a discussion on the list about the semantics of our
> dependencies. I'll try to get it going again.



Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
macports-dev mailing list
[email protected]
https://lists.macosforge.org/mailman/listinfo/macports-dev

Reply via email to