gmp's archive_site in Portfile disables custom archive_sites

Christopher Jones jonesc at hep.phy.cam.ac.uk
Thu Jan 5 23:16:38 UTC 2017


> On 5 Jan 2017, at 9:46 pm, Peter Brommer <p.brommer at warwick.ac.uk> wrote:
> 
> Hi Chris, 
> 
> thanks for your reply.
> 
>> On 5 Jan 2017, at 16:08, Chris Jones <jonesc at hep.phy.cam.ac.uk> wrote:
> 
>> Presumably the first machine, that creates the local binaries runs just fine, but then the second doesn't use these for this one port, and attempts to build again from source ? If correct, is this really a problem ? From the discussion in the ticket this is exactly what is wanted for this port, as it compiles very specifically for the machine it is to be used on. You *want* it to therefore build from source on each and every machine.
> 
> Yes, that indeed is not ideal, but probably necessary for this port. I don’t object to this.
> 
> The point is that after doing that, macports has “forgotten” about my local source for all further ports to upgrade and falls back to the default macports source. It’s not a big issue indeed (as cancelling and restarting the port upgrade outdated fixes it), but it might be a symptom of a bug, which I don’t have the energy to track down. 
> 
> Unfortunately, after I noticed this behaviour (not using my private archive site as first instance), I aborted the operation, before I got to a port where binaries are not available from macports. So I don’t know if it just got the priorities wrong (using default macports binaries before local archive), or had forgotten about my archive sites completely.


Ah right, I see the issue. I wouldn’t have guess that *all* ports would avoid the local tar balls….

One last question, just a thought, but what exactly are you aiming to gain from this setup ? most ports have binaries available direct from macports. I have multiple machines and have never really had any issues with just letting each do its own thing. Are you looking to save downloads, or CPU time building those ports without upstream binaries, or … ? Just seems to me a complicated setup for little return.

Chris

> 
> Sorry for not making this clearer initially, I hope you see my point now.
> 
> Peter
> 
> 
>> 
>> Chris
>> 
>> On 05/01/17 14:23, Peter Brommer wrote:
>>> Hi,
>>> 
>>> it appears that gmp @6.1.2_0 breaks my macports setup, due to its use of “archive_sites” in the Portfile (to avoid installing from precompiled binaries).
>>> 
>>> I have two Macs, and I use the setup described in https://trac.macports.org/wiki/howto/ShareArchives2 - one machine creates the local binaries not available from macports, the other one uses them via a customised archive_sites.conf.
>>> 
>>> gmp (and maybe other ports) unsets archive_sites to avoid installing from binaries (as discussed in https://trac.macports.org/ticket/41614), so it ignores my custom archive_site (aside: in principle, it should fail as I request binary-only with port -b upgrade, but if I read https://trac.macports.org/ticket/50885 correctly that will be addressed in a future macports release). However, after building, installing and activating gmp, macports no longer used my local archive_site, but falls back to the default archive_sites.
>>> 
>>> I’m not sure if this is a bug, and I have a hard time reproducing it, as the second time I run the command, gmp is already upgraded, and it keeps using my local archive_site. Should I still file this as a bug? And against what?
>>> 
>>> Cheers,
>>> 
>>> Peter
>>> 
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1910 bytes
Desc: not available
URL: <http://lists.macports.org/pipermail/macports-users/attachments/20170105/299427e5/attachment.bin>


More information about the macports-users mailing list