[MacPorts] #66288: Discussion re pre-build ports for PPC

[MacPorts]

#66288: Discussion re pre-build ports for PPC
-------------------------------------------+----------------------
 Reporter:  barracuda156                   |      Owner:  kencu
     Type:  enhancement                    |     Status:  assigned
 Priority:  Normal                         |  Milestone:
Component:  buildbot/mpbb                  |    Version:  2.8.0
 Keywords:  powerpc, leopard, snowleopard  |       Port:
-------------------------------------------+----------------------
 Ken, I was thinking a few days back about Macports and PPC, and I have few
 proposals to consider. Initially I thought to write a single comprehensive
 post, but now I deem it is more practical to separate associated, but not
 tied to each other issues.

 So here is question # 1.

 FWIU, we do not have PPC buildbots anymore (correct me if I am wrong).
 ''If you or someone can allocate one PowerMac to this end, it would be a
 superior solution to what I am suggesting''. If so, the text below is
 irrelevant.

 Second best solution, IMHO:

 I can upload ports that I build on PPC to MacPorts servers, which can be
 then used for archives. I am willing to do so, since a) that benefits PPC
 users, which otherwise may need ridiculous amount of time to build certain
 ports (forget Cmake or GCC, think Guile LOL), and b) marginal cost for
 myself is nearly zero (time to do uploads, which can be automated).

 Anticipating objections/doubts/questions, I will list obligations on my
 side, which I see necessary, as well as some negative conditions (that is,
 not from MacPorts side, but from my side: what I cannot or will not do).

 What I commit to, if this proposal is followed through:

 1. I will only submit ports built with unmodified portfiles from current
 MacPorts master, no exceptions (unless explicitly agreed to and approved
 prior to such submission).
 2. As long as I have physical access to my PPC hardware, I will update
 important (or agreed to upon) ports reasonably fast. Unlike the first
 point, this is not hard commitment at least for technical reasons, but I
 intend to keep this going.

 Then, limitations:

 1. Common-sensical, but anyway: is not feasible (or personally desirable
 either) to build all existing ports, so choice of what is built and
 submitted is at my discretion. Of course, if something is desirable to
 have for MacPorts, I am open to such suggestions and will try to fulfill
 those. (Buildbot automation cannot be implemented.)
 2. I am ready to support primarily Leopard (if 10.6 is considered at a
 later point, that is easy, but let leave this topic for now aside). I can
 do something for Tiger occasionally, but generally speaking, not
 particularly interested in it.
 3. For reasons you are well aware of, ppc64 and universal builds are often
 problematic. I will try to build few important ports for ppc+ppc64 (as
 long as they do build from MacPorts master), but no commitments here.

 For sure, ''there is no implication of any commitment from MacPorts
 side''. You do not have to use anything at all from pre-built ports from
 my side or may choose to use some and neglect another. Obviously, if I see
 that nothing at all is used, I shall interpret it as an end to the
 project.
 (As I said, marginal cost of try this for my is close to zero: whatever I
 build locally, I can upload, and do not care if a specific port is used or
 not.)

 The scope of ports considered can be narrowed down (like, only pre-build
 Guile, GCC and mpich), if that is preferable for MacPorts, or left open
 (i.e. whatever I build from standard portfiles, I upload).

 What do you think?

 (P. S. It is fine to drop the idea, if you or anyone thinks it is of no
 value for MacPorts.)

-- 
Ticket URL: <https://trac.macports.org/ticket/66288>
MacPorts <https://www.macports.org/>
Ports system for macOS