On Fri, Nov 13, 2009 at 1:36 PM, William Stein <wst...@gmail.com> wrote: > On Fri, Nov 13, 2009 at 8:14 AM, Bill Hart <goodwillh...@googlemail.com> > wrote: >> >> The compiler itself segfaulted. I don't see how that can be caused by >> MPIR. You aren't even linking with MPIR at that point. You are only >> including gmp.h whilst compiling something else. >> >> In order to build the MPIR test suite after building MPIR with >> --enable-fat, gmp.h must be included and compiled. So we know the MPIR >> code itself should be ok. > > I don't know what to tell you, except the following is a *fact* with > 32-bit Debian 5 (and other 32-bit Linuxes). (This doesn't happen on > 64-bit Debian 5). > > 1. Build Sage without --enable-fat (for MPIR) and everything is > fine; the full build completes and tests pass. > > 2. Build Sage with --enable-fat (for MPIR) and the compiler > segfaults later when building Pari. > >> The bug may be triggered when building Pari against some version of >> MPIR, but in this case it is far more likely the bug is elsewhere. >> Certainly there is a bug in gcc irregardless, as compilers are not >> supposed to segfault, no matter how broken the code is you are >> compiling (unless you have some broken environment variables hanging >> around). > > I agree that this is a bug in GCC, but unfortunately it is a bug in a > lot of deployed GCC's, and we (=the Sage project) don't have the > option of telling people to just "upgrade to GCC-4.4.2". > > Thanks for looking at this further (in your other messages). I just > hope to find a workaround because at least once a week (maybe more) we > specifically get complains from people about "Illegal Instruction" > errors when they try to use Sage binaries. I'm trying GMP-4.3.1 > with --enable-fat right now to see what happens.
OK, for completeness, in fact using GMP-4.3.1 with --enable-fat does work, i.e., PARI builds fine. I haven't finished building the rest of Sage though. So if the xgcd issues were fixed, this could be a temporary workaround for distribution until the MPIR-related issue is resolved. -- william --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en -~----------~----~----~----~------~----~------~--~---