I have placed a tarball here: http://sage.math.washington.edu/home/wbhart/mpir-0.9.0.tar.gz
It will become the final version of MPIR if no further build failures are reported. Testing has not occurred on SkyNet yet. I would appreciate volunteers who have access (./configure; make; make check is fine). Known issues: - MSYS and mingw32 --enable-fat build issue has not been resolved (will be fixed in service release) - cygwin32 does not support make check after build in a directory whose full path contains spaces - no 32 bit x86 support for Apple gcc (see http://trac.mpir.org/mpir_trac/ticket/94 for workaround) - make tune will core dump with small probability on some old Sun blade machines with Sun cc compiler - possible memory leak in t-locale (test code only and no one has replicated it since it was reported) Bill. On Jan 31, 4:26 pm, Bill Hart <goodwillh...@googlemail.com> wrote: > Actually, I just committed a fix for a bug which would definitely > cause at least the first few of the errors above. I had a screwed up > regular expression in configure.in (I used a very simplistic regexp > tutorial and it was wrong in this situation). > > Are you able to try again. > > Bill. > > 2009/1/31 Bill Hart <goodwillh...@googlemail.com>: > > > Hi Case, > > > thanks for checking this out. Can you send me a copy of the fat.h that > > it generates. Perhaps there is some clue. > > > Bill. > > > 2009/1/31 Case Vanhorsen <cas...@gmail.com>: > > >> On 1/30/09, Bill Hart <goodwillh...@googlemail.com> wrote: > > >>> 2009/1/30 Case Vanhorsen <cas...@gmail.com>: > > >>> > On Fri, Jan 30, 2009 at 4:07 AM, Bill Hart > >>> > <goodwillh...@googlemail.com> wrote: > > >>> >> 2009/1/30 Case Vanhorsen <cas...@gmail.com>: > > >>> >>> On 1/29/09, Bill Hart <goodwillh...@googlemail.com> wrote: > > >>> >>>> Case, > > >>> >>>> thank you, that would be very helpful. > > >>> >>>> I have just finished building and testing MPIR on cygwin. Apart from > >>> >>>> a > >>> >>>> warning about undefined symbols in the library (it didn't specify > >>> >>>> what > >>> >>>> they were), it built and passed make check just fine, even with > >>> >>>> --enable-fat. > > >>> >>> I was able to build successfully on mingw32. I still had to delete the > >>> >>> lines containing "mpn_preinv_add" and "mpn_preinv_lshift" from fat.h. > >>> >>> I did delete all the .asm files from the /mpn directory. Did I miss > >>> >>> something? > > >>> >> Ah, so we still have that issue. When you say you deleted *all* the > >>> >> .asm files, you don't mean you deleted every single .asm file in there > >>> >> do you? > > >>> > Yes. Then I ran ./configure --enable-fat and the required .asm files > >>> > were copied back into the directory. > > >>> Oh, that's fine. We'll eventually fix the make distclean issue which > >>> will rectify this. Actually make clean should probably clear these > >>> files. > > >>> >> In order to test a fix for this issue, could you try copying the files > >>> >> pre_divrem_1.c and pre_mod_1.c from mpn/x86_64/fat to mpn/x86/fat > >>> >> *before* running configure. Don't delete any .asm files and see if it > >>> >> builds. If not could you let me know what the error messages are. > >>> >> You'll have to do make clean before running make again, unless you > >>> >> check the whole repo out from scratch again. > > >> I get the following error: > > >> /bin/sh ../libtool --tag=CC --mode=compile gcc -std=gnu99 > >> -DHAVE_CONFIG_H -I. -I. -I.. -D__GMP_WITHIN_GMP -I.. -DOPERATION_`echo > >> fat | sed 's/_$//'` -m32 -O2 -fomit-frame-pointer -mno-cygwin -c -o > >> fat.lo fat.c > >> gcc -std=gnu99 -DHAVE_CONFIG_H -I. -I. -I.. -D__GMP_WITHIN_GMP -I.. > >> -DOPERATION_fat -m32 -O2 -fomit-frame-pointer -mno-cygwin -c fat.c -o > >> fat.o > >> In file included from ../gmp-impl.h:109, > >> from fat.c:31: > >> ../fat.h:432: warning: type defaults to `int' in declaration of > >> `DECL_preinv_add_n' > >> ../fat.h:432: warning: parameter names (without types) in function > >> declaration > >> ../fat.h:432: warning: data definition has no type or storage class > >> fat.c: In function `__gmpn_cpuvec_init': > >> fat.c:223: error: structure has no member named `preinv_add_n' > >> fat.c:223: error: `__gmpn_preinv_add_n_p6' undeclared (first use in > >> this function) > >> fat.c:223: error: (Each undeclared identifier is reported only once > >> fat.c:223: error: for each function it appears in.) > >> make[2]: *** [fat.lo] Error 1 > >> make[2]: Leaving directory `/c/svn/mpir/trunk/mpn' > >> make[1]: *** [all-recursive] Error 1 > >> make[1]: Leaving directory `/c/svn/mpir/trunk' > >> make: *** [all] Error 2 > > >> I will install cygwin and check if it's a mingw32 bug. > > >>> > I'll try this tonight. > > >>> Again, thanks. > > >>> Bill. > > >>> >> It's a puzzling error, as it doesn't happen on cygwin and I think my > >>> >> original diagnosis may have been wrong. > > >>> >>>> With regard to the demos issue, I *think* that this is due to use > >>> >>>> using svn for our repository. It doesn't preserve file properties, > >>> >>>> and > >>> >>>> one of the file properties for the relevant files indicates that it > >>> >>>> needs to be rebuilt, thus prompting it to try and run yacc. > > >>> >>>> Quite possibly we can avoid this by building MPIR and doing make > >>> >>>> distclean before making the release tarball. When I go to make the > >>> >>>> release, I'll let you know in advance about the tarball and perhaps > >>> >>>> you can verify whether the demos problem goes away. If not we'll have > >>> >>>> to disable building of demos until we can find a proper solution. > > >>> >>> I was able to work around the demos issue by using "touch *.c *.h" in > >>> >>> the demos/calc directory. If you do that before you create the > >>> >>> tarball, it should work. > > >>> >>>> Bill. > > >>> >>>> 2009/1/29 Case Vanhorsen <cas...@gmail.com>: > > >>> >>>> > Bill, > > >>> >>>> > I can test a mingw32 build this evening. I just need to work around > >>> >>>> > the demos issue. I'll compile on a few other platforms, too. > > >>> >>>> > Case > > >>> >>>> > On Thu, Jan 29, 2009 at 9:56 AM, Bill Hart > >>> >>>> > <goodwillh...@googlemail.com> wrote: > > >>> >>>> >> The fat binary support on x86_64 is now fixed. > > >>> >>>> >> I fixed the issue with fat binaries on x86_64 machines with ABI = > >>> >>>> >> 32. > >>> >>>> >> Hopefully this also fixes it for cygwin and mingw. But it takes > >>> >>>> >> over > >>> >>>> >> 1.5 hours to build under cygwin (make completes successfully), > >>> >>>> >> however > >>> >>>> >> I didn't get through make check because I forgot to build in a > >>> >>>> >> directory without spaces in the name (a known cygwin bug), so it > >>> >>>> >> crapped out. > > >>> >>>> >> Thus, I want to delay dealing with any further issues for fat > >>> >>>> >> binaries > >>> >>>> >> on mingw32 or cygwin until a service release, MPIR 0.9.1. > > >>> >>>> >> So can we try again for release? Anything broken for anyone? > > >>> >>>> >> Bill. --~--~---------~--~----~------------~-------~--~----~ 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 -~----------~----~----~----~------~----~------~--~---
