Re: GSoC Enquiry: Improving binary packaging

[Michael Goffioul]

On Thu, Jun 6, 2013 at 2:23 PM, John W. Eaton <address@hidden> wrote:

On 06/05/2013 12:14 PM, Michael Goffioul wrote:

On Wed, Jun 5, 2013 at 12:27 AM, Anirudha Bose <address@hidden

<mailto:address@hidden>> wrote:

    Hi.
    First of all thanks a lot for selecting me as a GSoC student. It was
    a real surprise for me. Looking forward to a very productive summer
    ahead! :)

    I was going through the list of accepted projects under GNU Project
    [1] when I saw that my mentor is Michael Goffioul. However the ideas
    page of Octave said that the mentor for this project will be either
    John W. Eaton or Jordi Gutiérrez Hermoso. Please clarify this and
    let me know.


Hi Anirudha,

I will be your mentor for this project. John and Jordi are already busy
mentoring other projects, so I offered my services for this one. You
might not know me, so let me briefly introduce myself. I've been part of
the octave project for a couple of years (since 2007 iirc) and have been
on and off, depending on the workload of my regular job. I worked on
various things like: Win32 native port of octave, java integration,
java-based graphics toolkit (JHandles), builtin octave graphics system,
OpenGL renderer, text renderer, and more recently classdef
implementation. I'm also maintaining MSVC-based binaries for Windows.

I'm an electronic engineer, but nowadays I'm mainly working in the
software engineering domain.

My timezone is UTC-4, so I'm not sure we'll be able to communicate
through IRC. Nevertheless, I advise you to hang out on IRC, as there are
a bunch of people that will be able to help you if I'm not around. My
IRC nickname is "goffioul".

One thing that I would like you to think about sooner rather than later
is how you will integrate your project with the current octave code
base. It doesn't mean it has to be part of octave source tree (for
instance jwe's MXE is a separate repository), but as some point we'll
need a plan to integrate your work into the octave "project".

What are the plans for this project?  I'd like to see us converge on one build system rather than duplicating effort (again).

I know I am guilty of duplicating previous work when I was trying to get Octave to build with a MinGW cross compiler because I didn't really look at what Michael and others had done in the past for building Octave on Windows, but I'd like to see us avoid further duplication of effort.

I agree that there should be a single build system for "exotic" platforms. And I see this project as a way to merge the existing efforts and converge towards a single system. The project description is on google-melange, but basically there are 2 items on the table: MXE and GUB. The goal is to select one and make it produce binaries for as many platforms as possible.

For MXE, what you've done so far is a good starting point. I understand it's able to cross-compile to Windows from Linux, but also native-compile under Linux and MinGW(?). Anirudha, in your project description, you're stating that MXE cannot be used for OS X. Why is it so? Couldn't jwe's version of MXE be re-used to cross-/native-compile octave?

Concerning GUB, I don't know it, so I can't tell. But if jwe's MXE can be used, that would be my first choice, in order to leverage all the work that's already been done.

Michael.