Hi,

On 03 Nov 2014, at 17:56 , Thomas Friedrichsmeier 
<[email protected]> wrote:
>> It seems like the port writes directly to the target location instead of
>> using the destroot directory. This usually happens if the port does not
>> respect the DESTDIR=... argument passed to 'make install'.
> 
> yes. Only it is what happened in the past. And now the files in question 
> are destrooted, and thus part of the image (as can also be seen in the 
> buildbot's log). And that's why port refuses to activate the image: Some 
> files 
> contained in it are already present at the target location (unregistered).

yep.


> The question at hand is how to get rid of the files that have been installed 
> without being registered _previously_.

Which is why I came up with the pre-activate phase for port R discussed earlier.


> P.S.: Some more clarification: I am not a regular mac or macports user 
> (pretty 
> much only for porting RKWard), and so I'm not up to speed on all this. 
> However, I am both upstream author, and the debian package maintainer of 
> rkward, and so I do know first-hand that rkward's "make install" is designed 
> to respect DESTDIR argument. Apparently there used to be a bug (now fixed) in 
> either cmake, or R, or something else, that caused the files to be mis-
> installed to the system path, anyway, or not registered for any other reason. 
> RKWard's upstream installation procedure has not changed in three years, with 
> respect to these files.

Well, if you say that that fix has happened 3 years ago, I am wondering why 
we’re
finding RKWard’s files not properly destrooted here…


Anyway, I guess I’ll not reimplement R’s pre-activate phase using fs-traverse as
it seems sufficient for me the way it is atm. I’ll file a ticket for R’s 
maintainer
once I am done with it. This way we should be able to fix the situation on the
buildbots (as well as on MacPorts installations of RKWard-users).

Greets,
Marko


P.S.: Sorry for the confusion caused on my end by not grasping at all times 
what’s
going on here! :-/ Perhaps I should have left all of this to R’s maintainer...

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

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

Reply via email to