Patches for old GMP releases

• [2005-06-01] Users of MPFR are advised not to use the version distributed with GMP, since that is obsolete and lacks several important bug fixes. Please download a newer release of MPFR from www.mpfr.org before contemplating to configure GMP with --enable-mpfr.
• [2005-05-23] On 64-bit systems, mpf output conversion yielding more than 2^31-1 digits will return only part of the digits. There are other problems with these routines when using the experimental nails feature. Patch
• [2005-04-08] There is a buffer overrun condition in the internal function mpn_rootrem, affecting the user-level functions mpz_root, mpz_rootrem and mpz_perfect_power_p. The potential overrun is a function of input operand size, i.e., with large operands the overrun can be large. Patch
• [2005-01-13] On 64-bit systems, mpz_sizeinbase returns garbage for operands greater than or equal to 2^31. The error is in gmp-impl.h, which you need to patch. (Please note that operands of this size might trigger other problems in GMP; we will address this thoroughly for GMP 5.)
• [2004-09-29] On AMD64 systems (Opteron, Athlon64, and Intel EM64T) compilation of the file mpz/n_pow_ui.c might fail with an error message. This is due to a very annoying change of the meaning of the "q" register flag in GCC 3.1. Patch for gmp-impl.h. Note: There is no need to apply this patch unless you run into this problem, and no known GCC version triggers this problem at the default (-O2) GMP optimization level.
• [2004-09-23] On HP-UX running on an Itanium processor, the underscore test fails too. The same workaround applies as to IRIX (see next item). (Notice that only the 32-bit ABI will have any chance of working on this platform.)
• [2004-04-30] On IRIX 5 and 6 under the old ABI=o32 mode using the native SGI tools, the GMP configure test for an assembler underscore prefix fails. Pass gmp_cv_asm_underscore=no to configure as a workaround.

• mpf_sub has two problems. When the result is zero the exponent may not be zeroed, causing incorrect operation in subsequent operations such as mpf_cmp_ui and mpz_set_f. When high limbs cancel out, zeros at the top of the remainder are incorrectly counted towards the destination precision, causing a loss of accuracy and in particular possibly leading to a zero result when the operands are not equal. Patch for mpf/sub.c.

• On AMD K6 systems (k6*-*-*), a bug in the assembler code may cause wrong results from mpz_gcd and other functions using GCDs. Patch for mpn/x86/k6/gcd_finda.asm.
• On FreeBSD with --enable-profiling, certain mpz routines may give incorrect results due to gcc (version 3.3 for instance) not realizing that the profiler calls may clobber regparm registers. If profiling is desired then change gmp-impl.h to force USE_LEADING_REGPARM to 0.
• On VxWorks, non-PowerPC targets may fail to build due to PowerPC assembler code being selected in longlong.h. For those targets, remove defined (PPC) from the two #if's in that file.
• On powerpc*-apple-darwin7.2.0, a bug in the C library sscanf affects certain EOF returns from gmp_sscanf, causing a failure in GMP tests/misc/t-scanf.c (line 1421). Hopefully Apple will fix this before too long. Past versions of Darwin have been ok.
• On MINGW, missing import/export directives may cause MPFR programs to fail to link, and may cause a (spurious) failure in tests/mpz/t-fib_ui. Add __GMP_DECLSPEC to mpn_copyi, mpn_copyd and __gmp_fib_table in gmp-impl.h.
• On perl 5.8.2, the demos/perl module may fail to accept substr() strings as arguments, due to lack of scalar magic handling in GMP.xs. For now the suggestion is to ignore the test.pl failures and to avoid passing substrings directly to the GMP module.
• On recent MINGW in a DLL build, programs using the low-level mpn routines may fail to link, with a message about "auto-import". MPFR is affected by this too. The cause is missing COFF type directives in the asm files. The current suggestion is to use a static build instead of a DLL.
• On Sparc V9 Solaris in 32-bit mode using Sun cc, the default flags don't include -xarch=v8plus needed for the GMP assembler code. Configure with ./configure CFLAGS="-xtarget=native -xarch=v8plus -xO4"
• demos/factorize.c can go into an almost-infinite loop for certain inputs. Patch.
• Pentium4 diveby3.asm contains a non-PIC data reference. On most systems this merely means that page won't be shared, but on Solaris it might provoke a linker error. Patch.
• mpfrxx.h has a typo in assignments of mpfr_class from mpf_class. Patch.
• On Opteron/AMD64 systems with GCC, count_trailing_zeros in longlong.h may cause errors from the assembler about incorrect registers used with bsfq. Patch.
• On MINGW with a DLL build of GMP, the demos/perl module fails to compile due to static variables initialized from DLL function pointers. The suggestion is to either use a static build of GMP, or compile the perl part as C++ using "make CC=g++" (in GMP.xs class_or_croak, rename the class variable since that's a keyword in C++).
• gmpxx.h / and % operators on mpz_class and long operand combinations may give incorrect results. Patch (applications must be recompiled).
• On ARM systems, mpfr/rnd_mode.c may fail to compile due to this chip not having all the rounding modes demanded (eg. FE_TOWARDZERO). Removing the offending cases will allow the library to build, though several test programs will fail.
• On S390 systems, the double format is not identified by gmp-impl.h, which can cause an MPFR build to fail. Patch.
• On longlong limb systems, mpz_gcd_ui may return a bad value when v==0 and u fits a limb but not a ulong. Patch.
• On some ARM GNU/Linux systems floating point format might not be detected, due to literal double constants being compiled incorrectly. This can cause an MPFR build to fail. Patch.

• When computing large-order roots using mpz_root or mpz_rootrem, an allocation for a temporary variable can become too small. This bug can result in incorrect results, but a buffer overrun like this can be a security risk in certain environments. Patch.
• The gmp printf family of functions outputs incorrect values when a number begins with more than 256 zeros. Typical behavior is that 1 to 255 extra zeros are output. In addition to giving incorrect results, this bug might result in buffer overruns. Patch.
• mpz_addmul and mpz_submul on Solaris 8 using a shared libgmp.so built by gcc 2.96 or higher may give incorrect results due to regparm registers being clobbered by lazy binding. This might arise on other i386 systems too, though a GNU system with GLIBC 2.1 or higher is known to be safe. Patch.
• On alphaev67 and alphaev68, longlong.h uses GCC style __asm__ blocks without checking that GCC is being used. Patch.
• HP-UX IA-64 systems have some problems. The assembly code and setups don't work, so it's necessary to use the generic C code with --host=none-hp-hpux11 (check ./config.guess for the exact OS type). Also double floats are not correctly treated as big-endian; in gmp-impl.h move the __ia64 test for ieee_double_extract down to the big-endian case. Also, the unbundled HP compilers coredump when compiling mpn/generic/mul_fft.c at high optimization levels; step into the mpn build directory and type "make CFLAGS=+O1 mul_fft.lo".
• The version of libtool currently in use rather aggressively strips compiler options when creating a shared library. This will hopefully be relaxed in the future, but for now if this is a problem the suggestion is to create a little script to hide them, for instance CC=gcc-with-my-options.

• On a 64-bit PowerPC under AIX, an ABI=32 build gets incorrect setups, resulting in asm files failing to build. In the generated config.m4 change include_mpn(`powerpc64/aix.m4') to include_mpn(`powerpc32/aix.m4')
• Libtool cannot build C++ shared libraries on MacOS X, so use --disable-shared when using --enable-cxx on that system.
• OpenBSD x86 shared libraries fail to build due to some assembly syntax. For this system, add the following line to mpn/x86/x86-defs.m4: define(_GLOBAL_OFFSET_TABLE_,__GLOBAL_OFFSET_TABLE_)
• Sparc V9 64-bit shared libraries may fail to build with GCC. In configure add gcc_64_ldflags="-Wc,-m64" just after the v9 gcc_64_cflags, or alternately override at make time with make GMP_LDFLAGS="-Wc,-m64".
• On alphaev67 and alphaev68 a shared library may fail to build due to __gmpn_clz_tab. Remove the #ifdef from mp_clz_tab.c.
• The tests/mpz/t-sizeinbase.c has been seen failing on HPPA systems, due to some rash assumptions made about pointer arithmetic in that program. This can be ignored.
• The native "make" on some versions of HP-UX complains about SCANF_OBJECTS not defined. This is due to the trailing backslash on the end of the preceding PRINTF_OBJECTS. Remove that from the generated Makefile.
• Sparc systems with GCC 3.1 are not identified as having IEEE floats. Patch.
• mpf_inp_str returns an incorrect character count for valid input, and leaks memory for invalid input. Patch. (This problem has been present since 2.0.)
• mpz_class::get_num and get_den return a special type which behaves like mpz_class& but is not identical. An actual mpz_class& can be obtained with this patch.
• mpz_powm_ui may return a result outside the range 0 to m-1 in certain circumstances (e==1, abs(b)>m, m normalized). Patch. (This problem was present also in 4.0.)
• Random routines may hit a broken assert in randraw.c. Patch. (This problem was introduced in 4.1.)
• Old Solaris systems don't recognise the .register directives in mpn/sparc64/*.asm. Delete those lines if necessary.
• CPU type alphaev56 incorrectly selects just the ev4 code. Configure as alphaev5 for a speedup.
• ./configure doesn't recognise the version number format used by GCC 3.1 on MINGW and Solaris, which can mean -march=pentiumpro is omitted unnecessarily. Patch.
• The test program t-scanf fails on x86_64 systems, due to some sloppy varargs handling in that program. The library code as such is believed to be ok.

• The C++ cmp functions described in the manual are in gmpxx.h as compare. The line __GMP_DEFINE_BINARY_TYPE_FUNCTION(int, compare, __gmp_cmp_function) can be edited to make it match the manual.
• gmp_printf and friends have a bug affecting %s and %p. In printf/doprnt.c under case 'p': case 's':, the break should be goto next.
• gmp_printf and friends are supposed to accept %zd etc for size_t output, but in printf/doprnt.c need a line case 'z': added above the set_type: label.
• gmp_printf, gmp_scanf, and friends, incorrectly claim size_t is unavailable when given %zn. In printf/doprnt.c and scanf/doscan.c remove the #if HAVE_SIZE_T conditionals, leaving just the code for the "true" case.
• On BeOS and IRIX 5 (and possibly elsewhere) intmax_t is in <inttypes.h>. Change printf/doprnt.c, printf/repl-vsnprintf.c and scanf/doscan.c to use

  #if HAVE_INTTYPES_H
  # include <inttypes.h>
  #else
  # if HAVE_STDINT_H
  #  include <stdint.h>
  # endif
  #endif
  

• The native "make" on some versions of SCO OpenServer 5 and IRIX 6.5 complains about a "syntax error" on the line libgmp.la: $(libgmp_la_OBJECTS) $(libgmp_la_DEPENDENCIES), apparently due to the length of the dependencies list. Either use GNU Make, or as a workaround remove $(libgmp_la_DEPENDENCIES) from that line (which will make the initial build work, but if any recompiling is done libgmp.la might not be rebuilt at the right time).

• On i386 solaris 2.8, gcc -march=pentiumpro or -march=athlon doesn't know to use the zany cmov syntax required by the native assembler. This can affect GMP. The suggestion is either to use GNU binutils, or configure GMP with that flag downgraded to -march=pentium, for example on a P-Pro, P-II or P-III ./configure CFLAGS="-g -O2 -fomit-frame-pointer -mcpu=pentiumpro -march=pentium".
• On sparc v9 solaris 2.7, GMP uses ABI=64 by default, but this will fail if the kernel is in 32-bit mode. Use ./configure ABI=32 instead (in the future GMP will fall back to that automatically). But note that ABI=64 will give better performance, so consider rebooting into 64-bit mode (with ok boot kernel/sparcv9/unix).
• gmpxx.h needs the #include <strstream.h> uncommented. Either that or delete the unused get_str2 function which depends on it.

• This patch upgrades 4.0 to 4.0.1: ftp://ftp.gnu.org/gnu/gmp/gmp-4.0-4.0.1.diff.gz

• This patch converts GMP 3.1 to GMP 3.1.1: gmp-3.1-3.1.1.diff.gz

• This patch fixes a division bug that can cause an internal assertion failure when dividing numbers of at least a few hundred digits: gmp-3.0.1.p1.diff
• This patch fixes a serious bug to mpz_powm that can cause it to return results equal to the mod argument, when the result should really be 0: gmp-3.0.1.p2.diff
• There are bugs in the new random number routines that can cause the randomness to be poor, or even crashes or memory corruption. Please upgrade to GMP 3.1.1 if you need to use GMP's new random number functions.
• configure may not automatically detect the correct CPU under cygwin and may guess i686 which is not a valid CPU. The suggested workaround is "./configure --host=pentiumpro-pc-cygwin32" (with whatever CPU is approriate in place of pentiumpro).
• OpenBSD 2.6 m4 has a bug in eval, making it unsuitable for asm file processing. Upgrade to OpenBSD 2.7 (or later), or use GNU m4.

• This patch converts GMP 3.0 to GMP 3.0.1: gmp-3.0-3.0.1.diff.gz
• All patches for GMP 3.0.1 also apply to GMP 3.0. Please first upgrade to GMP 3.0.1 and then apply any patches to that release from the list above.

• A typo in mpq_equal makes it fail if the first compared operand is negative. This patch corrects the problem. Joe Keane found the problem and wrote the patch.
• Several errors in mpf_set_str and mpf_get_str make these functions segfault, and possibly generate incorrect results. This also affects mpf_out_str, mpf_inp_str, and mpf_init_set_str. This gzip'ed patch fixes these errors.
• A bug in a test case makes `make check' fail using the N32 ABI under Irix6. Apply this patch to mpz/tests/t-powm_ui.c to fix the bug. (Note that the N32 ABI isn't really supported in GMP; you need to pass `-n32 -mips3 -D_LONG_LONG_LIMB' manually to make GMP build correctly.)
• The performance on SPARC is poor because of a typo that causes some of the assembly code to be omitted. Apply this patch to mpn/configure.in to fix this.
• A bug in mpz_invert makes it sometimes return a negative result, and sometimes not detect when an inverse does not exist. Apply this patch to mpz/invert.c to fix this.
• A bug in mpq_add and mpq_sub makes them work unreliably due to references outside of allocated memory. Apply this patch to the mpq subdirectory to fix this.
• A change to GCC (from version 2.8) makes GMP fail to build on IBM RS/6000 computers running AIX. You'll see complaints of unrecognized assembly instructions. Pass CFLAGS="-g -O2 -mcpu=power" or CFLAGS="-g -O2 -mcpu=powerpc" to `make' to work around this. (Choose the form that is appropriate for your system; if you're unsure which processor type you have, try running config.guess from the GMP top level directory.)
• A bug in mpz_probab_prime_p makes it work unreliably for numbers < 4. Apply this patch to mpz/pprime_p.c to fix the bug.
• Some bugs in mpf_set_q makes it get the sign wrong for negative numbers, and write to unallocated memory for other numbers. Apply this patch to mpf/set_q.c to fix the bugs.
• A fencepost error in mpn_divrem makes it occationally write outside of the numerator/remainder operand. Even if this function is called by several other GMP functions, only mpf_get_str, mpf_set_q, and mpf_set_str call it in a way that risk to trigger the bug. Apply this patch to mpn/generic/divrem.c to fix the bug.

Some GNU/Linux distributions have a buggy ar program that omits many of the files from libgmp.a. As a result libgmp.a becomes useless. The best solution is to get a newer GNU release of binutils, version 2.6 or later. As a workaround, you could also apply this patch to the top-level directory of GMP 2.0.2, if you don't want to upgrade your GNU/Linux tools. The patch was written by Alexander Zimmermann.