On Oct 13, 2009, at 6:22 PM, William Stein wrote: > On Tue, Oct 13, 2009 at 10:41 AM, David Kirkby <david.kir...@onetel.net > > wrote: >> >> 2009/10/13 William Stein <wst...@gmail.com>: >>> >>> On Tue, Oct 13, 2009 at 3:53 AM, David Kirkby <drkir...@gmail.com> >>> wrote: >>>> >>>> >>>> >>>> On Oct 12, 8:27 pm, William Stein <wst...@gmail.com> wrote: >>>>> On Mon, Oct 12, 2009 at 6:58 AM, Dr. David Kirkby >>>> >>>>>> Most of the problems in Sage are not at the shell level. >>>>> >>>>> Yes, but the problems that have been discussed so far in this >>>>> thread >>>>> are. Also, busybox was proposed as a way of dealing with the >>>>> problems that are at the shell level. Discussing compiler >>>>> issues is >>>>> totally orthogonal to the entire rest of this thread. >>>> >>>> Another issue to consider is whether asking Solaris users to use >>>> a GNU- >>>> like environment is likely to backfire, and them lose interest in >>>> helping in a port to Solaris. When I asked on comp.unix.solaris for >>>> some help porting Sage, someone contacted me and said he was >>>> interested, but only if Sun compilers were used, not GNU. When I >>>> pointed out that realistically the most sensible thing to do was to >>>> get Sage working with gcc relieably first, then use the Sun >>>> compilers, >>>> he was simply not interested. I was hoping to contact him again >>>> once >>>> >>>> http://trac.sagemath.org/sage_trac/ticket/6579 >>>> >>>> is resolved. Iit's not a fix I feel confident doing properly, so >>>> I'd >>>> rather someone else did that one. I believe is is the single most >>>> important fix of all those I've submitted. William is probably the >>>> best person, as he understands this bit of code, not me. >>>> >>>> I somewhat doubt the other person who contacted me would be >>>> interested >>>> in using a GNU shell environment, giving he reluctance to work with >>>> gcc! I believe once >>>> >>>> http://trac.sagemath.org/sage_trac/ticket/6579 >>>> >>>> is resolved, there is a good chance of getting some help from other >>>> Solaris users. Tell them that they must use a GNU shell, and not >>>> a Sun >>>> one is likely to have a detrimental effect. >>>> >>>> Couple that, with the fact the code would almost certainly get less >>>> testing on Solaris, and I see it a recipe to kill off a Solaris >>>> port, >>>> not improve one. >>>> >>>>> Using Python (or busybox or whatever) for spkg-install's is not >>>>> the >>>>> answer to *all* portability issues. But it is a very good answer >>>>> to >>>>> some of them. >>>>> >>>>> William >>>> >>>> Maybe, but maybe it would have the dead opposite effect to what you >>>> think. >>> >>> That's a great argument. Unfortunately, you can make exactly the >>> same >>> argument with respect to asking Windows/Linux/OS X developers to >>> use a >>> POSIX-only environment. "Another issue to consider is whether >>> asking >>> Windows/Linux/OS X users to use a POSIX-only environment is likely >>> to >>> backfire, and them lose interest in helping in a port (or >>> maintenance) >>> on Windows/Linux/OS X." >>> >>> William >> >> For Unix/Linux environments, perhaps a sensible solution could be >> summed up in two or perhaps three sentances. >> >> 1) Ensure you respect all environment variables. >> >> 2) Check it works on all supported operating systems using GCC as a >> compiler. >> >> 3) Check it works on Cygwin, since a port to this is planned. >> >> I personally do not think that is an unreasonable request. One of the >> main justifications for this is that bugs often show on one platform >> and not another, even though they are not portability issues. >> >> Dave > > Robert B.: >>> I think this is a reasonable thing to require when reviewing spkgs, >>> but much of the above doesn't even make sense for most >>> contributions. > > This thread is mainly about Gonzalo's proposal that we target > something like busybox (or my suggestion "python") instead of POSIX > standard shell usage. Somehow it is amazingly difficult to keep this > discussion on track! > > Definitely your 1-3 are definitely a good idea though... it's hard to > argue with them.
Ah, I got my threads crossed. I't still a bit unclear, are we talking about the scripts in sage/local/bin, or install scripts for the various packages, or both? I like the idea of moving towards using Python as the scripting/build coordinating language, but that might make using non-standard compliers and/or cross compiling even more difficult. Personally, I would be much more likely to work on a python script than a shell script. On the other hand, I'm no build guru, so as long as I can just type make (or ./configure make) and it just works I'll be happy. - Robert --~--~---------~--~----~------------~-------~--~----~ To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
