On Wed, Dec 24, 2008 at 1:48 PM, mabshoff <mabsh...@googlemail.com> wrote: > On Dec 24, 10:24 am, "William Stein" <wst...@gmail.com> wrote: >> The *right* solution should be that as soon as I try to install the >> spkg, it fails with an error that says "install 32-bit userspace" and >> some hint as to what the heck that means. > > This is ridiculous and just plainly not doable. Building gcc is an art > form and even upstream can be a giant pain to build. And believe me, I > know what I am talking about since I have spend way more time with gcc > than nearly everybody in the group here..
I 100% stand by my statement. We can and will give useful error messages for officially supported optional spkg's. >> I actually didn't try it before. OK, trying it now, and it fails in a >> second. The >> java compiler on sage.math is: >> ---- >> Eclipse Java Compiler v_774_R33x, 3.3.1 >> Copyright IBM Corp 2000, 2007. All rights reserved. >> --- >> >> A correct spkg that can only build with one specific compiler (say >> sun's) should just say so at start and exit gracefully if that java >> isn't available. > > Jmol doesn't even *run* with anything but the Java JDK. There is no > Java compiler besides the Sun one or other proprietary ones like the > one from IBM that passes the Java certification process. Just because > some linux distributions like Ubuntu ship gcj for whatever reason > doesn't mean it will work. But I certainly agree that this should be > experimental. The spkg install should exit gracefully if one doesn't have the right compiler installed. That's not difficult. I hate software that fails halfway through in confusing ways when it could immediately exit at the beginning. > >> >> I want the Sage "optional" packages to *all* install fine on every >> >> system that we officially support Sage on. >> >> > No chance this is going to work. >> >> 1. Just because you don't know how to do something doesn't mean it can't >> work. > > You are kidding me? Obviously we're misunderstanding each other here. I'm not claiming that every spkg should be *supported* to run on every OS. However, when one tries to install on a non-supported platform, the failure to install should be very graceful. > I see no reason to support every piece of shit > software package that is out there. Your statement has nothing to do with what I'm talking about. And, Sage's optional package repository is not "every piece of shit software package out there". > The vast majority of those > packages are riddled with build bugs and I am more than content fixing > build issues in the packages we have in the core. I'm not telling you or anybody else to do anything. I'm describing how things are going to be very soon with the optional spkg's in Sage. Period. Perhaps a word about my vision for Sage is necessary at this point. Sage is a way to bring together the mathematical research community and to work to improve the situation with respect to software that is available, especially open source software. I am a mathematical researcher, and I care greatly about the software resources that are available to my colleagues in research and teaching. I hope that makes sense. >> 2. If I listened to naysayers Sage wouldn't exist. > > That is not at all my point. I have done more for fixing build issue > in the last year than anybody else out there. This is true. Many many thanks! >> In any case, if I clearly define what supported OS means, and am >> carefully about what goes in the optional spkg repo, the above *will* >> work. This was always the intention. The only reason it isn't >> realized now is that nobody has seriously worked on it. > > Yes, but that means seriously pruning what is there. We will see. >> >> There should be a single >> >> command to easily install *all* of them, and another easy way to run >> >> all relevant optional doctests (that depend on all the free optional >> >> spkg's). >> >> > That won't work either. >> >> See 1 and 2 above :-) > > It cannot be done since there is binary only code in the repo. Or try > building phc on Itanium - even building an ada compiler on Itanium > involves cross compiling the toolchain and that is just plainly not > worth the effort. I spend some serious time on this and it is a giant > pain. You might be missing my point. It is one thing for a "binary only" package in the repo to silently "pretend to install" but not work at all (the current situation), or for a package to mysteriously break halfway through the install (often the case now), and another thing for an spkg to cleanly fail in a clear way at the beginning of the install, perhaps with the statement "this spkg cannot be installed on Itaniums". >> > All we can shoot for is to get them all >> > working on sage.math. There are also optional doctests in sage that >> > depend on experimental spkgs like M2. And M2 does not build on Solaris >> > as is for example. >> >> Bad example, since Sage doesn't build on Solaris as is. Solaris isn't >> an officially supported OS to run Sage on yet. > > So what, it will be soon enough. M2 doesn't build on FreeBSD 6 either. When the spkg is installed, it should check that the OS is supported. If not, it immediately fails with a very clear statement about the situation wrt that OS. This is not impossible. And it's something that can and will be automatically tested. William --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
