On Sat, Jun 19, 2010 at 1:29 PM, Dr. David Kirkby <david.kir...@onetel.net> wrote: > On 06/19/10 07:49 PM, William Stein wrote: >> >> On Fri, Jun 11, 2010 at 6:00 PM, Dr. David Kirkby >> <david.kir...@onetel.net> wrote: >>> >>> http://trac.sagemath.org/sage_trac/ticket/8049 >>> >>> was marked as a blocker by William, then marked as fixed. But there is >>> nothing there to indicate how it was fixed. Was the fix to manually copy >>> the >>> libgfortran library every time a binary distribution of Sage is made? I >>> assumed this would have been added to the sage-bdist script, but there is >>> nothing there to do it. >>> >> >> I use this script to build all the binaries I build: >> >> http://sage.math.washington.edu/home/wstein/bin/botdist >> >> You can re-open 8049 if you want to come up with some better fix, >> which works for everybody who might build binaries. It's complicated >> though, since some people might not want to copy libgfortran.so over. >> >> William >> > > I'll think about that one. > > I have a few ideas, but all of them take a bit of time to do. > > 1) The following will be more suitable for those that do not have gfortran > in /usr/bin. > > drkir...@hawk:~$ dirname `command -v gfortran` | sed > 's/bin/lib\/libgfortran.so/' > /usr/local/gcc-4.4.4/lib/libgfortran.so > > > Though that too would need changing for 64-bit Solaris on SPARC, and yet > again for 64-bit Solaris on x64. > > 2) It seems to me that sage-bdist needs to be modified to handle 3 different > requirements. > > 1) Copies no libraries as now. > 2) Copies libfortran.so > 3) Copies libfortran.so and the other gcc libraries (libgcc_s.so, > libstdc++.so etc).
I agree. > > It seems unfortunate if your own private script is the only one able to > create binaries that are very useful. Yes, definitely. People should feel free to take it and modify it of course. But point taken! > > How would something like the following seem? > > $ ./sage -bdist 4.4.2 > (no gcc libraries copies, so unchanged behavior) > $ ./sage -bdist 4.4.2 fortran > (copies gfortran library ) > $ ./sage -bidst 4.4.2 all > (copies all gcc related libraries) > That seems reasonable. > The other thing that would be useful is the ability to change the > compression format. There are very significant savings in using p7zip format > on Solaris - dropping the binaries from around 500 MB to around 300 MB. > > I guess none of this is high on my priority list - I'd like to see the port > to OpenSolaris done first! You never know, it *might* beat the Cygwin port, > though that seems unlikely. > > At least you have managed to build Sage on Cygwin - I've not managed to do > it at all on OpenSolaris. All the easy "SAGE64" issues are solved - now only > the more difficult problems remain:) > > About 90 of the 98 standard packages are installing, but that still leaves > about 7 non-trivial ones to solve! See > > http://trac.sagemath.org/sage_trac/ticket/9026 > > (singular, pynac, libfplll, R, the Sage library and ECL being some of them.) > Yep. And we'll see how long until I port Sage to the iPad. That's *really* challenging, e.g., nobody has ever built Fortran on there, (I doubt I will either)... William -- 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
