On Tuesday 08 September 2009 00:57:28 Bill Hart wrote: > 2009/9/7 Jason Moxham <[email protected]>: > > On Saturday 05 September 2009 16:30:38 Bill Hart wrote: > >> I'm happy to bump our compatibility number if that is that is required. > >> However we should be meticulous about ensuring we are compatible with > >> GMP 4.3.1. In particular we need to ensure we normalise our extended gcd > >> precisely the same way as them. We should also go through the > >> documentation and ensure every single function has exactly the same > >> prototype and definition. > >> We really need to verify carefully though that this is the problem John > >> is encountering. > > > > A quick look and it appears we are compatible with 4.3.1 , but not 4.4 > > mpz_inits(....,0) is one function we are missing, can Windows do this > > sort of function ? > > Have we verified that changing the version number does fix the gcc > > problem? > > Ah yeah varargs or something. I can't see that being a C99 thing as > the rest of GMP is not C99. > > >> The original intention was to never change the GMP version > >> fro 4.2.1 because that is the version of GMP which we forked. > > > > I would of thought the reason for version numbers is to see wether the > > lib is compatible with version x.y.z , not which version we forked from , > > as who would care about that? > > That's certainly the main use for version numbers. If people would > prefer us to update it, I'm happy to. > > Does anyone know if there is a "Right Thing To Do". >
Apparently the proper way is http://sources.redhat.com/autobook/autobook/autobook_91.html http://www.linuxshowcase.org/2000/2000papers/papers/browndavid/browndavid_html/ but I get the impression that most people rely on the version we set ie 1.2.1 or gmp's 4.3.1 etc > >> Bill. > >> > >> 2009/9/5 Jason Moxham <[email protected]> > >> > >> > On Saturday 05 September 2009 13:05:19 John Cremona wrote: > >> > > 2009/9/5 Minh Nguyen <[email protected]>: > >> > > > Hi John, > >> > > > > >> > > > On Sat, Sep 5, 2009 at 1:31 AM, John > >> > > > Cremona<[email protected]> wrote: > >> > > > > >> > > > <SNIP> > >> > > > > >> > > >> Now, (1) the installed version of gmp is 4.2.1: > >> > > >> j...@selmer%grep GMP_VERSION /usr/include/gmp.h > >> > > >> #define GMP_VERSION "4.2.1" > >> > > > > >> > > > One way (at least for me) to debug this problem is to compile Sage > >> > > > 4.1.1 and 4.1.2.alpha0 on the same machine without GMP installed > >> > > > system-wide. The system-wide installation of GMP might be done > >> > > > through, say, a Linux distribution's package manager. After that, > >> > > > remove the system-wide GMP and proceed with compiling the above > >> > > > two versions of Sage. I'll try this approach on the Ubuntu 9.04 > >> > > > machine in my office, and report the results later. > >> > > > >> > > I'll have a go, though this would involve disabling/hiding the > >> > > system gmp, and doesn't gcc use that? This machine is owned by me & > >> > > Bill Hart (so I would have to ask him before doing that). > >> > > > >> > > Here's what happened, in order: > >> > > 1. Machine installed late July, with ubuntu and gcc-4.3.3. > >> > > 2. Sage 4.1.1 builds fine with that. > >> > > 3. Bill finds that 4.3.3 is broken, in that it compilies his correct > >> > > source code into something which does not run properly (for details, > >> > > ask Bill). > >> > > 4. Bill obtains source for gcc-4.4.1, but finds it does not build > >> > > owing to gmp not being recent enough. > >> > > 5. John installs latest gmp. > >> > > 6. Bill successfully builds gcc-4.4.1, which is now the machine's > >> > > >> > default. > >> > > >> > > 7. John fails to build sage-4.1.2.alpha0. > >> > > > >> > > At some point I also installed mpir on this machine. So now we have > >> > > the following confusing situation: > >> > > > >> > > /usr/include/gmp.h is actually mpir 1.2.2 (aliasing gmp 4.2.1) > >> > > /usr/local/include/gmp.h is gmp 4.3.1 > >> > > >> > So does the installation of GCC-4.4 require a GMP >4.3 ? , if so then > >> > to get > >> > it to use MPIR all we need to do is bump our gmp compatibility version > >> > numbers from 4.2.1 to 4.3.1 , assuming we have implemented any new > >> > documented > >> > features. > >> > > >> > > I don't know where the original system gmp is (or even if there was > >> > > one) since the only other files o nthe system called gmp.h are in > >> > > sage-*/local directories. And of course there are various versions > >> > > of the library in /usr/lib and /usr/local/lib. > >> > > > >> > > John > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en -~----------~----~----~----~------~----~------~--~---
