* [Buildroot] [PATCH 0/5] toolchain rework, part 1
@ 2010-12-06 19:40 Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 1/5] gmp: make it a proper package and bump to version 5.0.1 Gustavo Zacarias
` (6 more replies)
0 siblings, 7 replies; 8+ messages in thread
From: Gustavo Zacarias @ 2010-12-06 19:40 UTC (permalink / raw)
To: buildroot
This is the first batch of patches for the internal toolchain rework.
In this first stage we convert gmp, mpfr and mpc to proper packages and
hook them up for use in the current toolchain.
We also fix a small bug in the package infra with regards to versioned patches
that'll be required to avoid duplication in the binutils/gcc migration.
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Buildroot] [PATCH 1/5] gmp: make it a proper package and bump to version 5.0.1
2010-12-06 19:40 [Buildroot] [PATCH 0/5] toolchain rework, part 1 Gustavo Zacarias
@ 2010-12-06 19:40 ` Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 2/5] php: libgmp was renamed to gmp Gustavo Zacarias
` (5 subsequent siblings)
6 siblings, 0 replies; 8+ messages in thread
From: Gustavo Zacarias @ 2010-12-06 19:40 UTC (permalink / raw)
To: buildroot
* Convert gmp to a proper autotargets package
* Bump to version 5.0.1
* Hook it up for binutils/gcc to use
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
---
package/gmp/Config.in | 12 +----
package/gmp/gmp.mk | 109 ++------------------------------------
toolchain/binutils/binutils.mk | 12 ++---
toolchain/gcc/gcc-uclibc-4.x.mk | 6 +-
4 files changed, 16 insertions(+), 123 deletions(-)
diff --git a/package/gmp/Config.in b/package/gmp/Config.in
index 74596f8..226e088 100644
--- a/package/gmp/Config.in
+++ b/package/gmp/Config.in
@@ -1,14 +1,6 @@
-config BR2_PACKAGE_LIBGMP
+config BR2_PACKAGE_GMP
bool "gmp"
help
GNU Multiple Precision Arithmetic Library.
- http://www.swox.com/gmp/
-
-
-config BR2_PACKAGE_LIBGMP_HEADERS
- bool "gmp headers for target"
- depends on BR2_PACKAGE_LIBGMP
- help
- Install the gmp.h for the target.
-
+ http://gmplib.org/
diff --git a/package/gmp/gmp.mk b/package/gmp/gmp.mk
index a3488b8..cdddca1 100644
--- a/package/gmp/gmp.mk
+++ b/package/gmp/gmp.mk
@@ -3,108 +3,11 @@
# gmp
#
#############################################################
-GMP_VERSION:=4.2.4
-GMP_SOURCE:=gmp-$(GMP_VERSION).tar.bz2
-GMP_SITE:=$(BR2_GNU_MIRROR)/gmp
-GMP_CAT:=$(BZCAT)
-GMP_DIR:=$(TOOLCHAIN_DIR)/gmp-$(GMP_VERSION)
-GMP_TARGET_DIR:=$(BUILD_DIR)/gmp-$(GMP_VERSION)
-GMP_BINARY:=libgmp$(LIBTGTEXT)
-GMP_HOST_BINARY:=libgmp$(HOST_LIBEXT)
-GMP_LIBVERSION:=3.4.4
-# this is a workaround for a bug in GMP, please see
-# http://gmplib.org/list-archives/gmp-devel/2006-April/000618.html
-ifeq ($(HOST_EXEEXT),.exe)
-GMP_CPP_FLAGS:=CPPFLAGS=-DDLL_EXPORT
-endif
+GMP_VERSION = 5.0.1
+GMP_SITE = $(BR2_GNU_MIRROR)/gmp
+GMP_SOURCE = gmp-$(GMP_VERSION).tar.bz2
+GMP_INSTALL_STAGING = YES
-$(DL_DIR)/$(GMP_SOURCE):
- $(call DOWNLOAD,$(GMP_SITE),$(GMP_SOURCE))
-
-libgmp-source: $(DL_DIR)/$(GMP_SOURCE)
-
-$(GMP_DIR)/.unpacked: $(DL_DIR)/$(GMP_SOURCE)
- $(GMP_CAT) $(DL_DIR)/$(GMP_SOURCE) | tar -C $(TOOLCHAIN_DIR) $(TAR_OPTIONS) -
- toolchain/patch-kernel.sh $(GMP_DIR) package/gmp/ \*.patch
- $(CONFIG_UPDATE) $(@D)
- touch $@
-
-$(GMP_TARGET_DIR)/.configured: $(GMP_DIR)/.unpacked
- mkdir -p $(GMP_TARGET_DIR)
- (cd $(GMP_TARGET_DIR); rm -rf config.cache; \
- $(TARGET_CONFIGURE_OPTS) \
- $(TARGET_CONFIGURE_ARGS) \
- $(GMP_CPP_FLAGS) \
- $(GMP_DIR)/configure $(QUIET) \
- --target=$(GNU_TARGET_NAME) \
- --host=$(GNU_TARGET_NAME) \
- --build=$(GNU_HOST_NAME) \
- --prefix=/usr \
- $(PREFERRED_LIB_FLAGS) \
- $(DISABLE_NLS) \
- )
- touch $@
-
-$(GMP_TARGET_DIR)/.libs/$(GMP_BINARY): $(GMP_TARGET_DIR)/.configured
- #$(MAKE) $(TARGET_CONFIGURE_OPTS) -C $(GMP_TARGET_DIR)
- $(MAKE) -C $(GMP_TARGET_DIR)
-
-$(STAGING_DIR)/usr/lib/$(GMP_BINARY): $(GMP_TARGET_DIR)/.libs/$(GMP_BINARY)
- $(MAKE) DESTDIR=$(STAGING_DIR) -C $(GMP_TARGET_DIR) install
- $(STRIPCMD) $(STRIP_STRIP_UNNEEDED) $(STAGING_DIR)/usr/lib/libgmp$(LIBTGTEXT)*
-
-$(TARGET_DIR)/usr/lib/libgmp.so $(TARGET_DIR)/usr/lib/libgmp.so.$(GMP_LIBVERSION) $(TARGET_DIR)/usr/lib/libgmp.a: $(STAGING_DIR)/usr/lib/$(GMP_BINARY)
- cp -dpf $(STAGING_DIR)/usr/lib/libgmp$(LIBTGTEXT)* $(TARGET_DIR)/usr/lib/
-ifeq ($(BR2_PACKAGE_LIBGMP_HEADERS),y)
- test -d $(TARGET_DIR)/usr/include || mkdir -p $(TARGET_DIR)/usr/include
- cp -dpf $(STAGING_DIR)/usr/include/gmp.h $(TARGET_DIR)/usr/include/
-endif
-
-libgmp: $(TARGET_DIR)/usr/lib/libgmp$(LIBTGTEXT)
-stage-libgmp: $(STAGING_DIR)/usr/lib/$(GMP_BINARY)
-
-libgmp-clean:
- rm -f $(TARGET_DIR)/usr/lib/libgmp.* $(TARGET_DIR)/usr/include/gmp.h \
- $(STAGING_DIR)/usr/lib/libgmp* $(STAGING_DIR)/usr/include/gmp.h
- -$(MAKE) -C $(GMP_TARGET_DIR) clean
-
-libgmp-dirclean:
- rm -rf $(GMP_TARGET_DIR) $(GMP_DIR)
-
-GMP_DIR2:=$(TOOLCHAIN_DIR)/gmp-$(GMP_VERSION)-host
-GMP_HOST_DIR:=$(TOOLCHAIN_DIR)/gmp
-$(GMP_DIR2)/.configured: $(GMP_DIR)/.unpacked
- mkdir -p $(GMP_DIR2)
- (cd $(GMP_DIR2); rm -rf config.cache; \
- $(HOST_CONFIGURE_OPTS) \
- $(GMP_CPP_FLAGS) \
- $(GMP_DIR)/configure $(QUIET) \
- --prefix="$(GMP_HOST_DIR)" \
- --build=$(GNU_HOST_NAME) \
- --host=$(GNU_HOST_NAME) \
- --disable-shared \
- --enable-static \
- $(DISABLE_NLS) \
- )
- touch $@
-
-$(GMP_HOST_DIR)/lib/libgmp$(HOST_LIBEXT): $(GMP_DIR2)/.configured
- $(MAKE) -C $(GMP_DIR2) install
-
-host-libgmp: $(GMP_HOST_DIR)/lib/$(GMP_HOST_BINARY)
-host-libgmp-source: libgmp-source
-host-libgmp-clean:
- rm -rf $(GMP_HOST_DIR)
- -$(MAKE) -C $(GMP_DIR2) clean
-host-libgmp-dirclean:
- rm -rf $(GMP_HOST_DIR) $(GMP_DIR2)
-
-#############################################################
-#
-# Toplevel Makefile options
-#
-#############################################################
-ifeq ($(BR2_PACKAGE_LIBGMP),y)
-TARGETS+=libgmp
-endif
+$(eval $(call AUTOTARGETS,package,gmp))
+$(eval $(call AUTOTARGETS,package,gmp,host))
diff --git a/toolchain/binutils/binutils.mk b/toolchain/binutils/binutils.mk
index ef9f644..327a0f6 100644
--- a/toolchain/binutils/binutils.mk
+++ b/toolchain/binutils/binutils.mk
@@ -16,17 +16,15 @@ endif
BINUTILS_HOST_PREREQ:=
BINUTILS_TARGET_PREREQ:=
-BINUTILS_HOST_PREREQ:=$(TOOLCHAIN_DIR)/gmp/lib/libgmp$(HOST_LIBEXT) \
- $(TOOLCHAIN_DIR)/mpfr/lib/libmpfr$(HOST_LIBEXT)
-HOST_SOURCE += host-libgmp-source host-libmpfr-source
+BINUTILS_HOST_PREREQ:=host-gmp $(TOOLCHAIN_DIR)/mpfr/lib/libmpfr$(HOST_LIBEXT)
+HOST_SOURCE += host-gmp-source host-libmpfr-source
-BINUTILS_TARGET_PREREQ:=$(TARGET_DIR)/usr/lib/libgmp$(LIBTGTEXT) \
- $(TARGET_DIR)/usr/lib/libmpfr$(LIBTGTEXT)
+BINUTILS_TARGET_PREREQ:=gmp $(TARGET_DIR)/usr/lib/libmpfr$(LIBTGTEXT)
-EXTRA_BINUTILS_CONFIG_OPTIONS+=--with-gmp="$(GMP_HOST_DIR)"
+EXTRA_BINUTILS_CONFIG_OPTIONS+=--with-gmp=$(HOST_DIR)/usr
EXTRA_BINUTILS_CONFIG_OPTIONS+=--with-mpfr="$(MPFR_HOST_DIR)"
-BINUTILS_TARGET_CONFIG_OPTIONS=--with-gmp="$(GMP_TARGET_DIR)"
+BINUTILS_TARGET_CONFIG_OPTIONS=--with-gmp=$(STAGING_DIR)/usr
BINUTILS_TARGET_CONFIG_OPTIONS+=--with-mpfr="$(MPFR_TARGET_DIR)"
ifeq ($(BR2_PACKAGE_LIBMPC),y)
diff --git a/toolchain/gcc/gcc-uclibc-4.x.mk b/toolchain/gcc/gcc-uclibc-4.x.mk
index af085f7..cbcaadb 100644
--- a/toolchain/gcc/gcc-uclibc-4.x.mk
+++ b/toolchain/gcc/gcc-uclibc-4.x.mk
@@ -106,9 +106,9 @@ ifeq ($(BR2_INSTALL_OBJC),y)
GCC_TARGET_LANGUAGES:=$(GCC_TARGET_LANGUAGES),objc
endif
-GCC_WITH_HOST_GMP=--with-gmp=$(GMP_HOST_DIR)
+GCC_WITH_HOST_GMP = --with-gmp=$(HOST_DIR)/usr
GCC_WITH_HOST_MPFR=--with-mpfr=$(MPFR_HOST_DIR)
-HOST_SOURCE += host-libgmp-source host-libmpfr-source
+HOST_SOURCE += host-gmp-source host-libmpfr-source
ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.)
GCC_WITH_HOST_MPC=--with-mpc=$(MPC_HOST_DIR)
@@ -119,7 +119,7 @@ ifeq ($(BR2_INSTALL_FORTRAN),y)
GCC_TARGET_LANGUAGES:=$(GCC_TARGET_LANGUAGES),fortran
#GCC_TARGET_PREREQ+=$(TARGET_DIR)/usr/lib/libmpfr.so $(TARGET_DIR)/usr/lib/libgmp.so
#GCC_STAGING_PREREQ+=$(TOOLCHAIN_DIR)/mpfr/lib/libmpfr.so
-GCC_WITH_TARGET_GMP=--with-gmp="$(GMP_TARGET_DIR)"
+GCC_WITH_TARGET_GMP = --with-gmp=$(STAGING_DIR)/usr
GCC_WITH_TARGET_MPFR=--with-mpfr="$(MPFR_TARGET_DIR)"
ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.)
GCC_WITH_TARGET_MPC=--with-mpc="$(MPC_TARGET_DIR)"
--
1.7.2.2
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [Buildroot] [PATCH 2/5] php: libgmp was renamed to gmp
2010-12-06 19:40 [Buildroot] [PATCH 0/5] toolchain rework, part 1 Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 1/5] gmp: make it a proper package and bump to version 5.0.1 Gustavo Zacarias
@ 2010-12-06 19:40 ` Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 3/5] mpfr: make it a proper package and bump to version 3.0.0-p8 Gustavo Zacarias
` (4 subsequent siblings)
6 siblings, 0 replies; 8+ messages in thread
From: Gustavo Zacarias @ 2010-12-06 19:40 UTC (permalink / raw)
To: buildroot
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
---
package/php/Config.ext | 2 +-
package/php/php.mk | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/package/php/Config.ext b/package/php/Config.ext
index 5c7f5ea..4f1614e 100644
--- a/package/php/Config.ext
+++ b/package/php/Config.ext
@@ -77,7 +77,7 @@ comment "gettext support requires a toolchain with WCHAR support"
config BR2_PACKAGE_PHP_EXT_GMP
bool "gmp"
- select BR2_PACKAGE_LIBGMP
+ select BR2_PACKAGE_GMP
help
GMP support
diff --git a/package/php/php.mk b/package/php/php.mk
index af4fa29..f3ed0f1 100644
--- a/package/php/php.mk
+++ b/package/php/php.mk
@@ -89,7 +89,7 @@ endif
ifeq ($(BR2_PACKAGE_PHP_EXT_GMP),y)
PHP_CONF_OPT += --with-gmp=$(STAGING_DIR)/usr
- PHP_DEPENDENCIES += libgmp
+ PHP_DEPENDENCIES += gmp
endif
ifeq ($(BR2_PACKAGE_PHP_EXT_JSON),y)
--
1.7.2.2
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [Buildroot] [PATCH 3/5] mpfr: make it a proper package and bump to version 3.0.0-p8
2010-12-06 19:40 [Buildroot] [PATCH 0/5] toolchain rework, part 1 Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 1/5] gmp: make it a proper package and bump to version 5.0.1 Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 2/5] php: libgmp was renamed to gmp Gustavo Zacarias
@ 2010-12-06 19:40 ` Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 4/5] mpc: make it a proper package Gustavo Zacarias
` (3 subsequent siblings)
6 siblings, 0 replies; 8+ messages in thread
From: Gustavo Zacarias @ 2010-12-06 19:40 UTC (permalink / raw)
To: buildroot
* Convert mpfr to a proper autotargets package
* Bump to version 3.0.0-p8
* Hook it up for binutils/gcc to use
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
---
package/mpfr/Config.in | 13 +-
package/mpfr/mpfr-3.0.0-p8.patch | 1950 ++++++++++++++++++++++++++++++++++++++
package/mpfr/mpfr.mk | 126 +---
toolchain/binutils/binutils.mk | 10 +-
toolchain/gcc/gcc-uclibc-4.x.mk | 6 +-
5 files changed, 1969 insertions(+), 136 deletions(-)
create mode 100644 package/mpfr/mpfr-3.0.0-p8.patch
diff --git a/package/mpfr/Config.in b/package/mpfr/Config.in
index 302f79e..9acb0b4 100644
--- a/package/mpfr/Config.in
+++ b/package/mpfr/Config.in
@@ -1,18 +1,9 @@
-config BR2_PACKAGE_LIBMPFR
+config BR2_PACKAGE_MPFR
bool "mpfr"
default y if BR2_INSTALL_FORTRAN
- select BR2_PACKAGE_LIBGMP
+ select BR2_PACKAGE_GMP
help
C library for multiple-precision floating-point computations
with exact rounding.
http://www.mpfr.org/
-
-
-config BR2_PACKAGE_LIBMPFR_HEADERS
- bool "mpfr headers for target"
- depends on BR2_PACKAGE_LIBMPFR
- select BR2_PACKAGE_LIBGMP_HEADERS
- help
- Install the mpfr headers for the target.
-
diff --git a/package/mpfr/mpfr-3.0.0-p8.patch b/package/mpfr/mpfr-3.0.0-p8.patch
new file mode 100644
index 0000000..e521b2e
--- /dev/null
+++ b/package/mpfr/mpfr-3.0.0-p8.patch
@@ -0,0 +1,1950 @@
+This is the cumulative patch available at http://www.mpfr.org/mpfr-3.0.0/
+It's recommended by upstream
+
+Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
+---
+diff -Naurd mpfr-3.0.0-a/PATCHES mpfr-3.0.0-b/PATCHES
+--- mpfr-3.0.0-a/PATCHES 2010-06-23 11:02:49.000000000 +0000
++++ mpfr-3.0.0-b/PATCHES 2010-06-23 11:03:36.000000000 +0000
+@@ -0,0 +1 @@
++mpfr_out_str
+diff -Naurd mpfr-3.0.0-a/VERSION mpfr-3.0.0-b/VERSION
+--- mpfr-3.0.0-a/VERSION 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/VERSION 2010-06-23 11:03:20.000000000 +0000
+@@ -1 +1 @@
+-3.0.0
++3.0.0-p1
+diff -Naurd mpfr-3.0.0-a/mpfr.h mpfr-3.0.0-b/mpfr.h
+--- mpfr-3.0.0-a/mpfr.h 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.h 2010-06-23 11:03:20.000000000 +0000
+@@ -27,7 +27,7 @@
+ #define MPFR_VERSION_MAJOR 3
+ #define MPFR_VERSION_MINOR 0
+ #define MPFR_VERSION_PATCHLEVEL 0
+-#define MPFR_VERSION_STRING "3.0.0"
++#define MPFR_VERSION_STRING "3.0.0-p1"
+
+ /* Macros dealing with MPFR VERSION */
+ #define MPFR_VERSION_NUM(a,b,c) (((a) << 16L) | ((b) << 8) | (c))
+diff -Naurd mpfr-3.0.0-a/mpfr.texi mpfr-3.0.0-b/mpfr.texi
+--- mpfr-3.0.0-a/mpfr.texi 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.texi 2010-06-23 11:03:12.000000000 +0000
+@@ -2050,7 +2050,7 @@
+ are printed. If @var{base} is greater than 10, @samp{@@} will be used
+ instead of @samp{e} as exponent delimiter.
+
+-Return the number of bytes written, or if an error occurred, return 0.
++Return the number of characters written, or if an error occurred, return 0.
+ @end deftypefun
+
+ @deftypefun size_t mpfr_inp_str (mpfr_t @var{rop}, FILE *@var{stream}, int @var{base}, mpfr_rnd_t @var{rnd})
+diff -Naurd mpfr-3.0.0-a/out_str.c mpfr-3.0.0-b/out_str.c
+--- mpfr-3.0.0-a/out_str.c 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/out_str.c 2010-06-23 11:03:12.000000000 +0000
+@@ -22,6 +22,16 @@
+
+ #include "mpfr-impl.h"
+
++/* Warning! S should not contain "%". */
++#define OUT_STR_RET(S) \
++ do \
++ { \
++ int r; \
++ r = fprintf (stream, (S)); \
++ return r < 0 ? 0 : r; \
++ } \
++ while (0)
++
+ size_t
+ mpfr_out_str (FILE *stream, int base, size_t n_digits, mpfr_srcptr op,
+ mpfr_rnd_t rnd_mode)
+@@ -29,6 +39,7 @@
+ char *s, *s0;
+ size_t l;
+ mpfr_exp_t e;
++ int err;
+
+ MPFR_ASSERTN (base >= 2 && base <= 62);
+
+@@ -36,37 +47,16 @@
+ if (stream == NULL)
+ stream = stdout;
+
+- if (MPFR_IS_NAN(op))
+- {
+- fprintf (stream, "@NaN@");
+- return 3;
+- }
+-
+- if (MPFR_IS_INF(op))
+- {
+- if (MPFR_SIGN(op) > 0)
+- {
+- fprintf (stream, "@Inf@");
+- return 3;
+- }
+- else
+- {
+- fprintf (stream, "- at Inf@");
+- return 4;
+- }
+- }
+-
+- if (MPFR_IS_ZERO(op))
++ if (MPFR_UNLIKELY (MPFR_IS_SINGULAR (op)))
+ {
+- if (MPFR_SIGN(op) > 0)
+- {
+- fprintf(stream, "0");
+- return 1;
+- }
++ if (MPFR_IS_NAN (op))
++ OUT_STR_RET ("@NaN@");
++ else if (MPFR_IS_INF (op))
++ OUT_STR_RET (MPFR_IS_POS (op) ? "@Inf@" : "- at Inf@");
+ else
+ {
+- fprintf(stream, "-0");
+- return 2;
++ MPFR_ASSERTD (MPFR_IS_ZERO (op));
++ OUT_STR_RET (MPFR_IS_POS (op) ? "0" : "-0");
+ }
+ }
+
+@@ -77,21 +67,31 @@
+
+ l = strlen (s) + 1; /* size of allocated block returned by mpfr_get_str
+ - may be incorrect, as only an upper bound? */
+- if (*s == '-')
+- fputc (*s++, stream);
+
+- /* outputs mantissa */
+- fputc (*s++, stream); e--; /* leading digit */
+- fputc ((unsigned char) MPFR_DECIMAL_POINT, stream);
+- fputs (s, stream); /* rest of mantissa */
++ /* outputs possible sign and significand */
++ err = (*s == '-' && fputc (*s++, stream) == EOF)
++ || fputc (*s++, stream) == EOF /* leading digit */
++ || fputc ((unsigned char) MPFR_DECIMAL_POINT, stream) == EOF
++ || fputs (s, stream) == EOF; /* trailing significand */
+ (*__gmp_free_func) (s0, l);
++ if (MPFR_UNLIKELY (err))
++ return 0;
++
++ e--; /* due to the leading digit */
+
+ /* outputs exponent */
+ if (e)
+ {
++ int r;
++
+ MPFR_ASSERTN(e >= LONG_MIN);
+ MPFR_ASSERTN(e <= LONG_MAX);
+- l += fprintf (stream, (base <= 10 ? "e%ld" : "@%ld"), (long) e);
++
++ r = fprintf (stream, (base <= 10 ? "e%ld" : "@%ld"), (long) e);
++ if (MPFR_UNLIKELY (r < 0))
++ return 0;
++
++ l += r;
+ }
+
+ return l;
+diff -Naurd mpfr-3.0.0-a/tests/tout_str.c mpfr-3.0.0-b/tests/tout_str.c
+--- mpfr-3.0.0-a/tests/tout_str.c 2010-06-10 11:00:13.000000000 +0000
++++ mpfr-3.0.0-b/tests/tout_str.c 2010-06-23 11:03:12.000000000 +0000
+@@ -46,22 +46,54 @@
+ special (void)
+ {
+ mpfr_t x;
++ unsigned int n;
+
+ mpfr_init (x);
+
+ mpfr_set_nan (x);
+- mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ n = mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ if (n != 5)
++ {
++ printf ("Error: mpfr_out_str (file, 10, 0, NaN, MPFR_RNDN) wrote %u "
++ "characters instead of 5.\n", n);
++ exit (1);
++ }
+
+ mpfr_set_inf (x, 1);
+- mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ n = mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ if (n != 5)
++ {
++ printf ("Error: mpfr_out_str (file, 10, 0, +Inf, MPFR_RNDN) wrote %u "
++ "characters instead of 5.\n", n);
++ exit (1);
++ }
+
+ mpfr_set_inf (x, -1);
+- mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ n = mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ if (n != 6)
++ {
++ printf ("Error: mpfr_out_str (file, 10, 0, -Inf, MPFR_RNDN) wrote %u "
++ "characters instead of 6.\n", n);
++ exit (1);
++ }
+
+ mpfr_set_ui (x, 0, MPFR_RNDN);
+- mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ n = mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ if (n != 1)
++ {
++ printf ("Error: mpfr_out_str (file, 10, 0, +0, MPFR_RNDN) wrote %u "
++ "characters instead of 1.\n", n);
++ exit (1);
++ }
++
+ mpfr_neg (x, x, MPFR_RNDN);
+- mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ n = mpfr_out_str (fout, 10, 0, x, MPFR_RNDN);
++ if (n != 2)
++ {
++ printf ("Error: mpfr_out_str (file, 10, 0, -0, MPFR_RNDN) wrote %u "
++ "characters instead of 2.\n", n);
++ exit (1);
++ }
+
+ mpfr_clear (x);
+ }
+diff -Naurd mpfr-3.0.0-a/version.c mpfr-3.0.0-b/version.c
+--- mpfr-3.0.0-a/version.c 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/version.c 2010-06-23 11:03:20.000000000 +0000
+@@ -25,5 +25,5 @@
+ const char *
+ mpfr_get_version (void)
+ {
+- return "3.0.0";
++ return "3.0.0-p1";
+ }
+diff -Naurd mpfr-3.0.0-a/Makefile.in mpfr-3.0.0-b/Makefile.in
+--- mpfr-3.0.0-a/Makefile.in 2010-06-10 11:00:52.000000000 +0000
++++ mpfr-3.0.0-b/Makefile.in 2010-06-10 11:00:52.000000000 +0000
+@@ -239,6 +239,7 @@
+ distuninstallcheck_listfiles = find . -type f -print
+ distcleancheck_listfiles = find . -type f -print
+ ACLOCAL = @ACLOCAL@
++ALLOCA = @ALLOCA@
+ AMTAR = @AMTAR@
+ AR = @AR@
+ AS = @AS@
+diff -Naurd mpfr-3.0.0-a/PATCHES mpfr-3.0.0-b/PATCHES
+--- mpfr-3.0.0-a/PATCHES 2010-06-23 11:03:36.000000000 +0000
++++ mpfr-3.0.0-b/PATCHES 2010-06-25 13:23:13.000000000 +0000
+@@ -0,0 +1 @@
++alloca
+diff -Naurd mpfr-3.0.0-a/VERSION mpfr-3.0.0-b/VERSION
+--- mpfr-3.0.0-a/VERSION 2010-06-23 11:03:20.000000000 +0000
++++ mpfr-3.0.0-b/VERSION 2010-06-25 13:23:13.000000000 +0000
+@@ -1 +1 @@
+-3.0.0-p1
++3.0.0-p2
+diff -Naurd mpfr-3.0.0-a/acinclude.m4 mpfr-3.0.0-b/acinclude.m4
+--- mpfr-3.0.0-a/acinclude.m4 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/acinclude.m4 2010-06-10 11:00:14.000000000 +0000
+@@ -59,6 +59,9 @@
+ dnl sys/fpu.h - MIPS specific
+ AC_CHECK_HEADERS([sys/time.h sys/fpu.h])
+
++dnl Check how to get `alloca'
++AC_FUNC_ALLOCA
++
+ dnl SIZE_MAX macro
+ gl_SIZE_MAX
+
+diff -Naurd mpfr-3.0.0-a/configure mpfr-3.0.0-b/configure
+--- mpfr-3.0.0-a/configure 2010-06-10 11:00:51.000000000 +0000
++++ mpfr-3.0.0-b/configure 2010-06-25 13:23:05.000000000 +0000
+@@ -783,6 +783,7 @@
+ OBJDUMP
+ DLLTOOL
+ AS
++ALLOCA
+ MPFR_LIBM
+ ANSI2KNR
+ U
+@@ -5622,6 +5623,197 @@
+ done
+
+
++# The Ultrix 4.2 mips builtin alloca declared by alloca.h only works
++# for constant arguments. Useless!
++{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for working alloca.h" >&5
++$as_echo_n "checking for working alloca.h... " >&6; }
++if test "${ac_cv_working_alloca_h+set}" = set; then :
++ $as_echo_n "(cached) " >&6
++else
++ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
++/* end confdefs.h. */
++#include <alloca.h>
++int
++main ()
++{
++char *p = (char *) alloca (2 * sizeof (int));
++ if (p) return 0;
++ ;
++ return 0;
++}
++_ACEOF
++if ac_fn_c_try_link "$LINENO"; then :
++ ac_cv_working_alloca_h=yes
++else
++ ac_cv_working_alloca_h=no
++fi
++rm -f core conftest.err conftest.$ac_objext \
++ conftest$ac_exeext conftest.$ac_ext
++fi
++{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_working_alloca_h" >&5
++$as_echo "$ac_cv_working_alloca_h" >&6; }
++if test $ac_cv_working_alloca_h = yes; then
++
++$as_echo "#define HAVE_ALLOCA_H 1" >>confdefs.h
++
++fi
++
++{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for alloca" >&5
++$as_echo_n "checking for alloca... " >&6; }
++if test "${ac_cv_func_alloca_works+set}" = set; then :
++ $as_echo_n "(cached) " >&6
++else
++ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
++/* end confdefs.h. */
++#ifdef __GNUC__
++# define alloca __builtin_alloca
++#else
++# ifdef _MSC_VER
++# include <malloc.h>
++# define alloca _alloca
++# else
++# ifdef HAVE_ALLOCA_H
++# include <alloca.h>
++# else
++# ifdef _AIX
++ #pragma alloca
++# else
++# ifndef alloca /* predefined by HP cc +Olibcalls */
++char *alloca ();
++# endif
++# endif
++# endif
++# endif
++#endif
++
++int
++main ()
++{
++char *p = (char *) alloca (1);
++ if (p) return 0;
++ ;
++ return 0;
++}
++_ACEOF
++if ac_fn_c_try_link "$LINENO"; then :
++ ac_cv_func_alloca_works=yes
++else
++ ac_cv_func_alloca_works=no
++fi
++rm -f core conftest.err conftest.$ac_objext \
++ conftest$ac_exeext conftest.$ac_ext
++fi
++{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_func_alloca_works" >&5
++$as_echo "$ac_cv_func_alloca_works" >&6; }
++
++if test $ac_cv_func_alloca_works = yes; then
++
++$as_echo "#define HAVE_ALLOCA 1" >>confdefs.h
++
++else
++ # The SVR3 libPW and SVR4 libucb both contain incompatible functions
++# that cause trouble. Some versions do not even contain alloca or
++# contain a buggy version. If you still want to use their alloca,
++# use ar to extract alloca.o from them instead of compiling alloca.c.
++
++ALLOCA=\${LIBOBJDIR}alloca.$ac_objext
++
++$as_echo "#define C_ALLOCA 1" >>confdefs.h
++
++
++{ $as_echo "$as_me:${as_lineno-$LINENO}: checking whether \`alloca.c' needs Cray hooks" >&5
++$as_echo_n "checking whether \`alloca.c' needs Cray hooks... " >&6; }
++if test "${ac_cv_os_cray+set}" = set; then :
++ $as_echo_n "(cached) " >&6
++else
++ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
++/* end confdefs.h. */
++#if defined CRAY && ! defined CRAY2
++webecray
++#else
++wenotbecray
++#endif
++
++_ACEOF
++if (eval "$ac_cpp conftest.$ac_ext") 2>&5 |
++ $EGREP "webecray" >/dev/null 2>&1; then :
++ ac_cv_os_cray=yes
++else
++ ac_cv_os_cray=no
++fi
++rm -f conftest*
++
++fi
++{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_os_cray" >&5
++$as_echo "$ac_cv_os_cray" >&6; }
++if test $ac_cv_os_cray = yes; then
++ for ac_func in _getb67 GETB67 getb67; do
++ as_ac_var=`$as_echo "ac_cv_func_$ac_func" | $as_tr_sh`
++ac_fn_c_check_func "$LINENO" "$ac_func" "$as_ac_var"
++eval as_val=\$$as_ac_var
++ if test "x$as_val" = x""yes; then :
++
++cat >>confdefs.h <<_ACEOF
++#define CRAY_STACKSEG_END $ac_func
++_ACEOF
++
++ break
++fi
++
++ done
++fi
++
++{ $as_echo "$as_me:${as_lineno-$LINENO}: checking stack direction for C alloca" >&5
++$as_echo_n "checking stack direction for C alloca... " >&6; }
++if test "${ac_cv_c_stack_direction+set}" = set; then :
++ $as_echo_n "(cached) " >&6
++else
++ if test "$cross_compiling" = yes; then :
++ ac_cv_c_stack_direction=0
++else
++ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
++/* end confdefs.h. */
++$ac_includes_default
++int
++find_stack_direction ()
++{
++ static char *addr = 0;
++ auto char dummy;
++ if (addr == 0)
++ {
++ addr = &dummy;
++ return find_stack_direction ();
++ }
++ else
++ return (&dummy > addr) ? 1 : -1;
++}
++
++int
++main ()
++{
++ return find_stack_direction () < 0;
++}
++_ACEOF
++if ac_fn_c_try_run "$LINENO"; then :
++ ac_cv_c_stack_direction=1
++else
++ ac_cv_c_stack_direction=-1
++fi
++rm -f core *.core core.conftest.* gmon.out bb.out conftest$ac_exeext \
++ conftest.$ac_objext conftest.beam conftest.$ac_ext
++fi
++
++fi
++{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_c_stack_direction" >&5
++$as_echo "$ac_cv_c_stack_direction" >&6; }
++cat >>confdefs.h <<_ACEOF
++#define STACK_DIRECTION $ac_cv_c_stack_direction
++_ACEOF
++
++
++fi
++
++
+
+ for ac_header in stdint.h
+ do :
+@@ -7564,13 +7756,13 @@
+ else
+ lt_cv_nm_interface="BSD nm"
+ echo "int some_variable = 0;" > conftest.$ac_ext
+- (eval echo "\"\$as_me:7567: $ac_compile\"" >&5)
++ (eval echo "\"\$as_me:7759: $ac_compile\"" >&5)
+ (eval "$ac_compile" 2>conftest.err)
+ cat conftest.err >&5
+- (eval echo "\"\$as_me:7570: $NM \\\"conftest.$ac_objext\\\"\"" >&5)
++ (eval echo "\"\$as_me:7762: $NM \\\"conftest.$ac_objext\\\"\"" >&5)
+ (eval "$NM \"conftest.$ac_objext\"" 2>conftest.err > conftest.out)
+ cat conftest.err >&5
+- (eval echo "\"\$as_me:7573: output\"" >&5)
++ (eval echo "\"\$as_me:7765: output\"" >&5)
+ cat conftest.out >&5
+ if $GREP 'External.*some_variable' conftest.out > /dev/null; then
+ lt_cv_nm_interface="MS dumpbin"
+@@ -8772,7 +8964,7 @@
+ ;;
+ *-*-irix6*)
+ # Find out which ABI we are using.
+- echo '#line 8775 "configure"' > conftest.$ac_ext
++ echo '#line 8967 "configure"' > conftest.$ac_ext
+ if { { eval echo "\"\$as_me\":${as_lineno-$LINENO}: \"$ac_compile\""; } >&5
+ (eval $ac_compile) 2>&5
+ ac_status=$?
+@@ -10032,11 +10224,11 @@
+ -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
+ -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
+ -e 's:$: $lt_compiler_flag:'`
+- (eval echo "\"\$as_me:10035: $lt_compile\"" >&5)
++ (eval echo "\"\$as_me:10227: $lt_compile\"" >&5)
+ (eval "$lt_compile" 2>conftest.err)
+ ac_status=$?
+ cat conftest.err >&5
+- echo "$as_me:10039: \$? = $ac_status" >&5
++ echo "$as_me:10231: \$? = $ac_status" >&5
+ if (exit $ac_status) && test -s "$ac_outfile"; then
+ # The compiler can only warn and ignore the option if not recognized
+ # So say no if there are warnings other than the usual output.
+@@ -10371,11 +10563,11 @@
+ -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
+ -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
+ -e 's:$: $lt_compiler_flag:'`
+- (eval echo "\"\$as_me:10374: $lt_compile\"" >&5)
++ (eval echo "\"\$as_me:10566: $lt_compile\"" >&5)
+ (eval "$lt_compile" 2>conftest.err)
+ ac_status=$?
+ cat conftest.err >&5
+- echo "$as_me:10378: \$? = $ac_status" >&5
++ echo "$as_me:10570: \$? = $ac_status" >&5
+ if (exit $ac_status) && test -s "$ac_outfile"; then
+ # The compiler can only warn and ignore the option if not recognized
+ # So say no if there are warnings other than the usual output.
+@@ -10476,11 +10668,11 @@
+ -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
+ -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
+ -e 's:$: $lt_compiler_flag:'`
+- (eval echo "\"\$as_me:10479: $lt_compile\"" >&5)
++ (eval echo "\"\$as_me:10671: $lt_compile\"" >&5)
+ (eval "$lt_compile" 2>out/conftest.err)
+ ac_status=$?
+ cat out/conftest.err >&5
+- echo "$as_me:10483: \$? = $ac_status" >&5
++ echo "$as_me:10675: \$? = $ac_status" >&5
+ if (exit $ac_status) && test -s out/conftest2.$ac_objext
+ then
+ # The compiler can only warn and ignore the option if not recognized
+@@ -10531,11 +10723,11 @@
+ -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
+ -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
+ -e 's:$: $lt_compiler_flag:'`
+- (eval echo "\"\$as_me:10534: $lt_compile\"" >&5)
++ (eval echo "\"\$as_me:10726: $lt_compile\"" >&5)
+ (eval "$lt_compile" 2>out/conftest.err)
+ ac_status=$?
+ cat out/conftest.err >&5
+- echo "$as_me:10538: \$? = $ac_status" >&5
++ echo "$as_me:10730: \$? = $ac_status" >&5
+ if (exit $ac_status) && test -s out/conftest2.$ac_objext
+ then
+ # The compiler can only warn and ignore the option if not recognized
+@@ -12915,7 +13107,7 @@
+ lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
+ lt_status=$lt_dlunknown
+ cat > conftest.$ac_ext <<_LT_EOF
+-#line 12918 "configure"
++#line 13110 "configure"
+ #include "confdefs.h"
+
+ #if HAVE_DLFCN_H
+@@ -13011,7 +13203,7 @@
+ lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
+ lt_status=$lt_dlunknown
+ cat > conftest.$ac_ext <<_LT_EOF
+-#line 13014 "configure"
++#line 13206 "configure"
+ #include "confdefs.h"
+
+ #if HAVE_DLFCN_H
+diff -Naurd mpfr-3.0.0-a/mpfr.h mpfr-3.0.0-b/mpfr.h
+--- mpfr-3.0.0-a/mpfr.h 2010-06-23 11:03:20.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.h 2010-06-25 13:23:13.000000000 +0000
+@@ -27,7 +27,7 @@
+ #define MPFR_VERSION_MAJOR 3
+ #define MPFR_VERSION_MINOR 0
+ #define MPFR_VERSION_PATCHLEVEL 0
+-#define MPFR_VERSION_STRING "3.0.0-p1"
++#define MPFR_VERSION_STRING "3.0.0-p2"
+
+ /* Macros dealing with MPFR VERSION */
+ #define MPFR_VERSION_NUM(a,b,c) (((a) << 16L) | ((b) << 8) | (c))
+diff -Naurd mpfr-3.0.0-a/tests/Makefile.in mpfr-3.0.0-b/tests/Makefile.in
+--- mpfr-3.0.0-a/tests/Makefile.in 2010-06-10 11:00:52.000000000 +0000
++++ mpfr-3.0.0-b/tests/Makefile.in 2010-06-10 11:00:52.000000000 +0000
+@@ -960,6 +960,7 @@
+ red=; grn=; lgn=; blu=; std=
+ DISTFILES = $(DIST_COMMON) $(DIST_SOURCES) $(TEXINFOS) $(EXTRA_DIST)
+ ACLOCAL = @ACLOCAL@
++ALLOCA = @ALLOCA@
+ AMTAR = @AMTAR@
+ AR = @AR@
+ AS = @AS@
+diff -Naurd mpfr-3.0.0-a/version.c mpfr-3.0.0-b/version.c
+--- mpfr-3.0.0-a/version.c 2010-06-23 11:03:20.000000000 +0000
++++ mpfr-3.0.0-b/version.c 2010-06-25 13:23:13.000000000 +0000
+@@ -25,5 +25,5 @@
+ const char *
+ mpfr_get_version (void)
+ {
+- return "3.0.0-p1";
++ return "3.0.0-p2";
+ }
+diff -Naurd mpfr-3.0.0-a/PATCHES mpfr-3.0.0-b/PATCHES
+--- mpfr-3.0.0-a/PATCHES 2010-07-10 00:11:19.000000000 +0000
++++ mpfr-3.0.0-b/PATCHES 2010-07-10 00:12:50.000000000 +0000
+@@ -0,0 +1 @@
++gamma_underflow
+diff -Naurd mpfr-3.0.0-a/VERSION mpfr-3.0.0-b/VERSION
+--- mpfr-3.0.0-a/VERSION 2010-06-25 13:23:13.000000000 +0000
++++ mpfr-3.0.0-b/VERSION 2010-07-10 00:11:53.000000000 +0000
+@@ -1 +1 @@
+-3.0.0-p2
++3.0.0-p3
+diff -Naurd mpfr-3.0.0-a/gamma.c mpfr-3.0.0-b/gamma.c
+--- mpfr-3.0.0-a/gamma.c 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/gamma.c 2010-07-10 00:11:46.000000000 +0000
+@@ -274,7 +274,7 @@
+ /* we want an upper bound for x * [log(2-x)-1].
+ since x < 0, we need a lower bound on log(2-x) */
+ mpfr_ui_sub (xp, 2, x, MPFR_RNDD);
+- mpfr_log (xp, xp, MPFR_RNDD);
++ mpfr_log2 (xp, xp, MPFR_RNDD);
+ mpfr_sub_ui (xp, xp, 1, MPFR_RNDD);
+ mpfr_mul (xp, xp, x, MPFR_RNDU);
+
+@@ -303,8 +303,8 @@
+ {
+ mpfr_sub (tmp, tmp, tmp2, MPFR_RNDZ); /* low bnd on |sin(Pi*(2-x))| */
+ mpfr_ui_div (tmp, 12, tmp, MPFR_RNDU); /* upper bound */
+- mpfr_log (tmp, tmp, MPFR_RNDU);
+- mpfr_add (tmp, tmp, xp, MPFR_RNDU);
++ mpfr_log2 (tmp, tmp, MPFR_RNDU);
++ mpfr_add (xp, tmp, xp, MPFR_RNDU);
+ underflow = mpfr_cmp_si (xp, expo.saved_emin - 2) <= 0;
+ }
+
+diff -Naurd mpfr-3.0.0-a/mpfr.h mpfr-3.0.0-b/mpfr.h
+--- mpfr-3.0.0-a/mpfr.h 2010-06-25 13:23:13.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.h 2010-07-10 00:11:53.000000000 +0000
+@@ -27,7 +27,7 @@
+ #define MPFR_VERSION_MAJOR 3
+ #define MPFR_VERSION_MINOR 0
+ #define MPFR_VERSION_PATCHLEVEL 0
+-#define MPFR_VERSION_STRING "3.0.0-p2"
++#define MPFR_VERSION_STRING "3.0.0-p3"
+
+ /* Macros dealing with MPFR VERSION */
+ #define MPFR_VERSION_NUM(a,b,c) (((a) << 16L) | ((b) << 8) | (c))
+diff -Naurd mpfr-3.0.0-a/tests/tgamma.c mpfr-3.0.0-b/tests/tgamma.c
+--- mpfr-3.0.0-a/tests/tgamma.c 2010-06-10 11:00:13.000000000 +0000
++++ mpfr-3.0.0-b/tests/tgamma.c 2010-07-10 00:11:46.000000000 +0000
+@@ -461,6 +461,20 @@
+ mpfr_clear (x);
+ }
+
++/* bug found by Stathis, only occurs on 32-bit machines */
++static void
++test20100709 (void)
++{
++ mpfr_t x;
++ int inex;
++
++ mpfr_init2 (x, 100);
++ mpfr_set_str (x, "-4.6308260837372266e+07", 10, MPFR_RNDN);
++ inex = mpfr_gamma (x, x, MPFR_RNDN);
++ MPFR_ASSERTN(MPFR_IS_ZERO(x) && MPFR_IS_NEG(x) && inex > 0);
++ mpfr_clear (x);
++}
++
+ int
+ main (int argc, char *argv[])
+ {
+@@ -471,6 +485,7 @@
+ test_generic (2, 100, 2);
+ gamma_integer ();
+ test20071231 ();
++ test20100709 ();
+
+ data_check ("data/gamma", mpfr_gamma, "mpfr_gamma");
+
+diff -Naurd mpfr-3.0.0-a/version.c mpfr-3.0.0-b/version.c
+--- mpfr-3.0.0-a/version.c 2010-06-25 13:23:13.000000000 +0000
++++ mpfr-3.0.0-b/version.c 2010-07-10 00:11:53.000000000 +0000
+@@ -25,5 +25,5 @@
+ const char *
+ mpfr_get_version (void)
+ {
+- return "3.0.0-p2";
++ return "3.0.0-p3";
+ }
+diff -Naurd mpfr-3.0.0-a/PATCHES mpfr-3.0.0-b/PATCHES
+--- mpfr-3.0.0-a/PATCHES 2010-09-07 08:44:01.000000000 +0000
++++ mpfr-3.0.0-b/PATCHES 2010-09-07 08:48:46.000000000 +0000
+@@ -0,0 +1 @@
++mpfr_cmp/set_ui/si
+diff -Naurd mpfr-3.0.0-a/VERSION mpfr-3.0.0-b/VERSION
+--- mpfr-3.0.0-a/VERSION 2010-07-10 00:11:53.000000000 +0000
++++ mpfr-3.0.0-b/VERSION 2010-09-07 08:46:06.000000000 +0000
+@@ -1 +1 @@
+-3.0.0-p3
++3.0.0-p4
+diff -Naurd mpfr-3.0.0-a/mpfr.h mpfr-3.0.0-b/mpfr.h
+--- mpfr-3.0.0-a/mpfr.h 2010-07-10 00:11:53.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.h 2010-09-07 08:46:06.000000000 +0000
+@@ -27,7 +27,7 @@
+ #define MPFR_VERSION_MAJOR 3
+ #define MPFR_VERSION_MINOR 0
+ #define MPFR_VERSION_PATCHLEVEL 0
+-#define MPFR_VERSION_STRING "3.0.0-p3"
++#define MPFR_VERSION_STRING "3.0.0-p4"
+
+ /* Macros dealing with MPFR VERSION */
+ #define MPFR_VERSION_NUM(a,b,c) (((a) << 16L) | ((b) << 8) | (c))
+@@ -798,35 +798,45 @@
+ anyway. Checking with other ICC versions is needed. Possibly detect
+ whether warnings are produced or not with a configure test.
+ + Remove C++ too, since it complains too much. */
++/* Added casts to improve robustness in case of undefined behavior and
++ compiler extensions based on UB (in particular -fwrapv). MPFR doesn't
++ use such extensions, but these macros will be used by 3rd-party code,
++ where such extensions may be required.
++ Moreover casts to unsigned long have been added to avoid warnings in
++ programs that use MPFR and are compiled with -Wconversion; such casts
++ are OK since if X is a constant expression, then (unsigned long) X is
++ also a constant expression, so that the optimizations still work. */
+ #if defined (__GNUC__) && !defined(__ICC) && !defined(__cplusplus)
+ #if (__GNUC__ >= 2)
+ #undef mpfr_cmp_ui
+-/* We use the fact that mpfr_sgn on NaN sets the erange flag and returns 0. */
+-#define mpfr_cmp_ui(_f,_u) \
+- (__builtin_constant_p (_u) && (_u) == 0 ? \
+- mpfr_sgn (_f) : \
+- mpfr_cmp_ui_2exp ((_f),(_u),0))
++/* We use the fact that mpfr_sgn on NaN sets the erange flag and returns 0.
++ But warning! mpfr_sgn is specified as a macro in the API, thus the macro
++ mustn't be used if side effects are possible, like here. */
++#define mpfr_cmp_ui(_f,_u) \
++ (__builtin_constant_p (_u) && (unsigned long) (_u) == 0 ? \
++ (mpfr_sgn) (_f) : \
++ mpfr_cmp_ui_2exp ((_f), (unsigned long) (_u), 0))
+ #undef mpfr_cmp_si
+-#define mpfr_cmp_si(_f,_s) \
+- (__builtin_constant_p (_s) && (_s) >= 0 ? \
+- mpfr_cmp_ui ((_f), (_s)) : \
+- mpfr_cmp_si_2exp ((_f), (_s), 0))
++#define mpfr_cmp_si(_f,_s) \
++ (__builtin_constant_p (_s) && (long) (_s) >= 0 ? \
++ mpfr_cmp_ui ((_f), (unsigned long) (long) (_s)) : \
++ mpfr_cmp_si_2exp ((_f), (long) (_s), 0))
+ #if __GNUC__ > 2 || __GNUC_MINOR__ >= 95
+ #undef mpfr_set_ui
+-#define mpfr_set_ui(_f,_u,_r) \
+- (__builtin_constant_p (_u) && (_u) == 0 ? \
+- __extension__ ({ \
+- mpfr_ptr _p = (_f); \
+- _p->_mpfr_sign = 1; \
+- _p->_mpfr_exp = __MPFR_EXP_ZERO; \
+- (void) (_r); 0; }) : \
+- mpfr_set_ui_2exp ((_f), (_u), 0, (_r)))
++#define mpfr_set_ui(_f,_u,_r) \
++ (__builtin_constant_p (_u) && (unsigned long) (_u) == 0 ? \
++ __extension__ ({ \
++ mpfr_ptr _p = (_f); \
++ _p->_mpfr_sign = 1; \
++ _p->_mpfr_exp = __MPFR_EXP_ZERO; \
++ (void) (_r); 0; }) : \
++ mpfr_set_ui_2exp ((_f), (unsigned long) (_u), 0, (_r)))
+ #endif
+ #undef mpfr_set_si
+-#define mpfr_set_si(_f,_s,_r) \
+- (__builtin_constant_p (_s) && (_s) >= 0 ? \
+- mpfr_set_ui ((_f), (_s), (_r)) : \
+- mpfr_set_si_2exp ((_f), (_s), 0, (_r)))
++#define mpfr_set_si(_f,_s,_r) \
++ (__builtin_constant_p (_s) && (long) (_s) >= 0 ? \
++ mpfr_set_ui ((_f), (unsigned long) (long) (_s), (_r)) : \
++ mpfr_set_si_2exp ((_f), (long) (_s), 0, (_r)))
+ #endif
+ #endif
+
+diff -Naurd mpfr-3.0.0-a/tests/tcmp_ui.c mpfr-3.0.0-b/tests/tcmp_ui.c
+--- mpfr-3.0.0-a/tests/tcmp_ui.c 2010-06-10 11:00:13.000000000 +0000
++++ mpfr-3.0.0-b/tests/tcmp_ui.c 2010-09-07 08:45:12.000000000 +0000
+@@ -88,6 +88,126 @@
+ mpfr_clear (x);
+ }
+
++/* Since mpfr_cmp_ui and mpfr_cmp_si are also implemented by a macro
++ with __builtin_constant_p for GCC, check that side effects are
++ handled correctly. */
++static void
++check_macros (void)
++{
++ mpfr_t x;
++ int c;
++
++ mpfr_init2 (x, 32);
++
++ c = 0;
++ mpfr_set_ui (x, 17, MPFR_RNDN);
++ if (mpfr_cmp_ui (x, 17) != 0)
++ {
++ printf ("Error 1 on mpfr_cmp_ui(x,17) in check_macros\n");
++ exit (1);
++ }
++ if (mpfr_cmp_ui (x, (c++, 17)) != 0)
++ {
++ printf ("Error 2 on mpfr_cmp_ui(x,17) in check_macros\n");
++ exit (1);
++ }
++ if (c != 1)
++ {
++ printf ("Error 3 on mpfr_cmp_ui(x,17) in check_macros\n"
++ "(c = %d instead of 1)\n", c);
++ exit (1);
++ }
++ if (mpfr_cmp_si (x, 17) != 0)
++ {
++ printf ("Error 1 on mpfr_cmp_si(x,17) in check_macros\n");
++ exit (1);
++ }
++ if (mpfr_cmp_si (x, (c++, 17)) != 0)
++ {
++ printf ("Error 2 on mpfr_cmp_si(x,17) in check_macros\n");
++ exit (1);
++ }
++ if (c != 2)
++ {
++ printf ("Error 3 on mpfr_cmp_si(x,17) in check_macros\n"
++ "(c = %d instead of 2)\n", c);
++ exit (1);
++ }
++
++ c = 0;
++ mpfr_set_ui (x, 0, MPFR_RNDN);
++ if (mpfr_cmp_ui (x, 0) != 0)
++ {
++ printf ("Error 1 on mpfr_cmp_ui(x,0) in check_macros\n");
++ exit (1);
++ }
++ if (mpfr_cmp_ui (x, (c++, 0)) != 0)
++ {
++ printf ("Error 2 on mpfr_cmp_ui(x,0) in check_macros\n");
++ exit (1);
++ }
++ if (c != 1)
++ {
++ printf ("Error 3 on mpfr_cmp_ui(x,0) in check_macros\n"
++ "(c = %d instead of 1)\n", c);
++ exit (1);
++ }
++ if (mpfr_cmp_si (x, 0) != 0)
++ {
++ printf ("Error 1 on mpfr_cmp_si(x,0) in check_macros\n");
++ exit (1);
++ }
++ if (mpfr_cmp_si (x, (c++, 0)) != 0)
++ {
++ printf ("Error 2 on mpfr_cmp_si(x,0) in check_macros\n");
++ exit (1);
++ }
++ if (c != 2)
++ {
++ printf ("Error 3 on mpfr_cmp_si(x,0) in check_macros\n"
++ "(c = %d instead of 2)\n", c);
++ exit (1);
++ }
++
++ mpfr_clear (x);
++}
++
++/* Bug in r7114 */
++static void
++test_macros (void)
++{
++ mpfr_t x[3];
++ mpfr_ptr p;
++
++ mpfr_inits (x[0], x[1], x[2], (mpfr_ptr) 0);
++ mpfr_set_ui (x[0], 0, MPFR_RNDN);
++ p = x[0];
++ if (mpfr_cmp_ui (p++, 0) != 0)
++ {
++ printf ("Error in mpfr_cmp_ui macro: result should be 0.\n");
++ exit (1);
++ }
++ if (p != x[1])
++ {
++ printf ("Error in mpfr_cmp_ui macro: p - x[0] = %d (expecting 1)\n",
++ (int) (p - x[0]));
++ exit (1);
++ }
++ p = x[0];
++ if (mpfr_cmp_si (p++, 0) != 0)
++ {
++ printf ("Error in mpfr_cmp_si macro: result should be 0.\n");
++ exit (1);
++ }
++ if (p != x[1])
++ {
++ printf ("Error in mpfr_cmp_si macro: p - x[0] = %d (expecting 1)\n",
++ (int) (p - x[0]));
++ exit (1);
++ }
++ mpfr_clears (x[0], x[1], x[2], (mpfr_ptr) 0);
++}
++
+ int
+ main (void)
+ {
+@@ -216,6 +336,8 @@
+ mpfr_clear (x);
+
+ check_nan ();
++ check_macros ();
++ test_macros ();
+
+ tests_end_mpfr ();
+ return 0;
+diff -Naurd mpfr-3.0.0-a/version.c mpfr-3.0.0-b/version.c
+--- mpfr-3.0.0-a/version.c 2010-07-10 00:11:53.000000000 +0000
++++ mpfr-3.0.0-b/version.c 2010-09-07 08:46:06.000000000 +0000
+@@ -25,5 +25,5 @@
+ const char *
+ mpfr_get_version (void)
+ {
+- return "3.0.0-p3";
++ return "3.0.0-p4";
+ }
+diff -Naurd mpfr-3.0.0-a/PATCHES mpfr-3.0.0-b/PATCHES
+--- mpfr-3.0.0-a/PATCHES 2010-10-21 20:28:38.000000000 +0000
++++ mpfr-3.0.0-b/PATCHES 2010-10-21 20:28:38.000000000 +0000
+@@ -0,0 +1 @@
++tcan_round
+diff -Naurd mpfr-3.0.0-a/VERSION mpfr-3.0.0-b/VERSION
+--- mpfr-3.0.0-a/VERSION 2010-09-07 08:46:06.000000000 +0000
++++ mpfr-3.0.0-b/VERSION 2010-10-21 20:28:38.000000000 +0000
+@@ -1 +1 @@
+-3.0.0-p4
++3.0.0-p5
+diff -Naurd mpfr-3.0.0-a/mpfr.h mpfr-3.0.0-b/mpfr.h
+--- mpfr-3.0.0-a/mpfr.h 2010-09-07 08:46:06.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.h 2010-10-21 20:28:38.000000000 +0000
+@@ -27,7 +27,7 @@
+ #define MPFR_VERSION_MAJOR 3
+ #define MPFR_VERSION_MINOR 0
+ #define MPFR_VERSION_PATCHLEVEL 0
+-#define MPFR_VERSION_STRING "3.0.0-p4"
++#define MPFR_VERSION_STRING "3.0.0-p5"
+
+ /* Macros dealing with MPFR VERSION */
+ #define MPFR_VERSION_NUM(a,b,c) (((a) << 16L) | ((b) << 8) | (c))
+diff -Naurd mpfr-3.0.0-a/tests/tcan_round.c mpfr-3.0.0-b/tests/tcan_round.c
+--- mpfr-3.0.0-a/tests/tcan_round.c 2010-06-10 11:00:13.000000000 +0000
++++ mpfr-3.0.0-b/tests/tcan_round.c 2010-10-21 20:28:38.000000000 +0000
+@@ -41,7 +41,7 @@
+ /* avoid mpn_random which leaks memory */
+ for (i = 0; i < n; i++)
+ buf[i] = randlimb ();
+- p = (mpfr_prec_t) randlimb() % ((n-1) * GMP_NUMB_BITS) + MPFR_PREC_MIN;
++ p = randlimb() % ((n-1) * GMP_NUMB_BITS) + MPFR_PREC_MIN;
+ err = p + randlimb () % GMP_NUMB_BITS;
+ r1 = mpfr_round_p (buf, n, err, p);
+ r2 = mpfr_can_round_raw (buf, n, MPFR_SIGN_POS, err,
+diff -Naurd mpfr-3.0.0-a/version.c mpfr-3.0.0-b/version.c
+--- mpfr-3.0.0-a/version.c 2010-09-07 08:46:06.000000000 +0000
++++ mpfr-3.0.0-b/version.c 2010-10-21 20:28:38.000000000 +0000
+@@ -25,5 +25,5 @@
+ const char *
+ mpfr_get_version (void)
+ {
+- return "3.0.0-p4";
++ return "3.0.0-p5";
+ }
+diff -Naurd mpfr-3.0.0-a/PATCHES mpfr-3.0.0-b/PATCHES
+--- mpfr-3.0.0-a/PATCHES 2010-10-21 20:59:32.000000000 +0000
++++ mpfr-3.0.0-b/PATCHES 2010-10-21 20:59:32.000000000 +0000
+@@ -0,0 +1 @@
++mpfr_sub1
+diff -Naurd mpfr-3.0.0-a/VERSION mpfr-3.0.0-b/VERSION
+--- mpfr-3.0.0-a/VERSION 2010-10-21 20:28:38.000000000 +0000
++++ mpfr-3.0.0-b/VERSION 2010-10-21 20:59:32.000000000 +0000
+@@ -1 +1 @@
+-3.0.0-p5
++3.0.0-p6
+diff -Naurd mpfr-3.0.0-a/mpfr.h mpfr-3.0.0-b/mpfr.h
+--- mpfr-3.0.0-a/mpfr.h 2010-10-21 20:28:38.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.h 2010-10-21 20:59:32.000000000 +0000
+@@ -27,7 +27,7 @@
+ #define MPFR_VERSION_MAJOR 3
+ #define MPFR_VERSION_MINOR 0
+ #define MPFR_VERSION_PATCHLEVEL 0
+-#define MPFR_VERSION_STRING "3.0.0-p5"
++#define MPFR_VERSION_STRING "3.0.0-p6"
+
+ /* Macros dealing with MPFR VERSION */
+ #define MPFR_VERSION_NUM(a,b,c) (((a) << 16L) | ((b) << 8) | (c))
+diff -Naurd mpfr-3.0.0-a/sub1.c mpfr-3.0.0-b/sub1.c
+--- mpfr-3.0.0-a/sub1.c 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/sub1.c 2010-10-21 20:59:32.000000000 +0000
+@@ -37,7 +37,9 @@
+ mp_size_t cancel2, an, bn, cn, cn0;
+ mp_limb_t *ap, *bp, *cp;
+ mp_limb_t carry, bb, cc, borrow = 0;
+- int inexact, shift_b, shift_c, is_exact = 1, down = 0, add_exp = 0;
++ int inexact, shift_b, shift_c, add_exp = 0;
++ int cmp_low = 0; /* used for rounding to nearest: 0 if low(b) = low(c),
++ negative if low(b) < low(c), positive if low(b)>low(c) */
+ int sh, k;
+ MPFR_TMP_DECL(marker);
+
+@@ -196,7 +198,8 @@
+ }
+
+ #ifdef DEBUG
+- printf ("shift_b=%d shift_c=%d diffexp=%lu\n", shift_b, shift_c,
++ printf ("rnd=%s shift_b=%d shift_c=%d diffexp=%lu\n",
++ mpfr_print_rnd_mode (rnd_mode), shift_b, shift_c,
+ (unsigned long) diff_exp);
+ #endif
+
+@@ -307,17 +310,18 @@
+ {
+ if (MPFR_LIKELY(sh))
+ {
+- is_exact = (carry == 0);
+ /* can decide except when carry = 2^(sh-1) [middle]
+ or carry = 0 [truncate, but cannot decide inexact flag] */
+- down = (carry < (MPFR_LIMB_ONE << (sh - 1)));
+ if (carry > (MPFR_LIMB_ONE << (sh - 1)))
+ goto add_one_ulp;
+- else if ((0 < carry) && down)
++ else if ((0 < carry) && (carry < (MPFR_LIMB_ONE << (sh - 1))))
+ {
+ inexact = -1; /* result if smaller than exact value */
+ goto truncate;
+ }
++ /* now carry = 2^(sh-1), in which case cmp_low=2,
++ or carry = 0, in which case cmp_low=0 */
++ cmp_low = (carry == 0) ? 0 : 2;
+ }
+ }
+ else /* directed rounding: set rnd_mode to RNDZ iff toward zero */
+@@ -344,12 +348,32 @@
+ cn -= (long int) an + cancel2;
+
+ #ifdef DEBUG
+- printf ("last %d bits from a are %lu, bn=%ld, cn=%ld\n",
++ printf ("last sh=%d bits from a are %lu, bn=%ld, cn=%ld\n",
+ sh, (unsigned long) carry, (long) bn, (long) cn);
+ #endif
+
++ /* for rounding to nearest, we couldn't conclude up to here in the following
++ cases:
++ 1. sh = 0, then cmp_low=0: we can either truncate, subtract one ulp
++ or add one ulp: -1 ulp < low(b)-low(c) < 1 ulp
++ 2. sh > 0 but the low sh bits from high(b)-high(c) equal 2^(sh-1):
++ -0.5 ulp <= -1/2^sh < low(b)-low(c)-0.5 < 1/2^sh <= 0.5 ulp
++ we can't decide the rounding, in that case cmp_low=2:
++ either we truncate and flag=-1, or we add one ulp and flag=1
++ 3. the low sh>0 bits from high(b)-high(c) equal 0: we know we have to
++ truncate but we can't decide the ternary value, here cmp_low=0:
++ -0.5 ulp <= -1/2^sh < low(b)-low(c) < 1/2^sh <= 0.5 ulp
++ we always truncate and inexact can be any of -1,0,1
++ */
++
++ /* note: here cn might exceed cn0, in which case we consider a zero limb */
+ for (k = 0; (bn > 0) || (cn > 0); k = 1)
+ {
++ /* if cmp_low < 0, we know low(b) - low(c) < 0
++ if cmp_low > 0, we know low(b) - low(c) > 0
++ (more precisely if cmp_low = 2, low(b) - low(c) = 0.5 ulp so far)
++ if cmp_low = 0, so far low(b) - low(c) = 0 */
++
+ /* get next limbs */
+ bb = (bn > 0) ? bp[--bn] : 0;
+ if ((cn > 0) && (cn-- <= cn0))
+@@ -357,76 +381,115 @@
+ else
+ cc = 0;
+
+- /* down is set when low(b) < low(c) */
+- if (down == 0)
+- down = (bb < cc);
++ /* cmp_low compares low(b) and low(c) */
++ if (cmp_low == 0) /* case 1 or 3 */
++ cmp_low = (bb < cc) ? -2+k : (bb > cc) ? 1 : 0;
++
++ /* Case 1 for k=0 splits into 7 subcases:
++ 1a: bb > cc + half
++ 1b: bb = cc + half
++ 1c: 0 < bb - cc < half
++ 1d: bb = cc
++ 1e: -half < bb - cc < 0
++ 1f: bb - cc = -half
++ 1g: bb - cc < -half
++
++ Case 2 splits into 3 subcases:
++ 2a: bb > cc
++ 2b: bb = cc
++ 2c: bb < cc
++
++ Case 3 splits into 3 subcases:
++ 3a: bb > cc
++ 3b: bb = cc
++ 3c: bb < cc
++ */
+
+ /* the case rounding to nearest with sh=0 is special since one couldn't
+ subtract above 1/2 ulp in the trailing limb of the result */
+- if ((rnd_mode == MPFR_RNDN) && sh == 0 && k == 0)
++ if (rnd_mode == MPFR_RNDN && sh == 0 && k == 0) /* case 1 for k=0 */
+ {
+ mp_limb_t half = MPFR_LIMB_HIGHBIT;
+
+- is_exact = (bb == cc);
+-
+ /* add one ulp if bb > cc + half
+ truncate if cc - half < bb < cc + half
+ sub one ulp if bb < cc - half
+ */
+
+- if (down)
++ if (cmp_low < 0) /* bb < cc: -1 ulp < low(b) - low(c) < 0,
++ cases 1e, 1f and 1g */
+ {
+ if (cc >= half)
+ cc -= half;
+- else
++ else /* since bb < cc < half, bb+half < 2*half */
+ bb += half;
++ /* now we have bb < cc + half:
++ we have to subtract one ulp if bb < cc,
++ and truncate if bb > cc */
+ }
+- else /* bb >= cc */
++ else if (cmp_low >= 0) /* bb >= cc, cases 1a to 1d */
+ {
+ if (cc < half)
+ cc += half;
+- else
++ else /* since bb >= cc >= half, bb - half >= 0 */
+ bb -= half;
++ /* now we have bb > cc - half: we have to add one ulp if bb > cc,
++ and truncate if bb < cc */
++ if (cmp_low > 0)
++ cmp_low = 2;
+ }
+ }
+
+ #ifdef DEBUG
+- printf (" bb=%lu cc=%lu down=%d is_exact=%d\n",
+- (unsigned long) bb, (unsigned long) cc, down, is_exact);
++ printf ("k=%u bb=%lu cc=%lu cmp_low=%d\n", k,
++ (unsigned long) bb, (unsigned long) cc, cmp_low);
+ #endif
+- if (bb < cc)
++ if (cmp_low < 0) /* low(b) - low(c) < 0: either truncate or subtract
++ one ulp */
+ {
+ if (rnd_mode == MPFR_RNDZ)
+- goto sub_one_ulp;
++ goto sub_one_ulp; /* set inexact=-1 */
+ else if (rnd_mode != MPFR_RNDN) /* round away */
+ {
+ inexact = 1;
+ goto truncate;
+ }
+- else /* round to nearest: special case here since for sh=k=0
+- bb = bb0 - MPFR_LIMB_HIGHBIT */
++ else /* round to nearest */
+ {
+- if (is_exact && sh == 0)
+- {
+- /* For k=0 we can't decide exactness since it may depend
+- from low order bits.
+- For k=1, the first low limbs matched: low(b)-low(c)<0. */
+- if (k)
+- {
+- inexact = 1;
+- goto truncate;
+- }
+- }
+- else if (down && sh == 0)
+- goto sub_one_ulp;
+- else
+- {
+- inexact = (is_exact) ? 1 : -1;
++ /* If cmp_low < 0 and bb > cc, then -0.5 ulp < low(b)-low(c) < 0,
++ whatever the value of sh.
++ If sh>0, then cmp_low < 0 implies that the initial neglected
++ sh bits were 0 (otherwise cmp_low=2 initially), thus the
++ weight of the new bits is less than 0.5 ulp too.
++ If k > 0 (and sh=0) this means that either the first neglected
++ limbs bb and cc were equal (thus cmp_low was 0 for k=0),
++ or we had bb - cc = -0.5 ulp or 0.5 ulp.
++ The last case is not possible here since we would have
++ cmp_low > 0 which is sticky.
++ In the first case (where we have cmp_low = -1), we truncate,
++ whereas in the 2nd case we have cmp_low = -2 and we subtract
++ one ulp.
++ */
++ if (bb > cc || sh > 0 || cmp_low == -1)
++ { /* -0.5 ulp < low(b)-low(c) < 0,
++ bb > cc corresponds to cases 1e and 1f1
++ sh > 0 corresponds to cases 3c and 3b3
++ cmp_low = -1 corresponds to case 1d3 (also 3b3) */
++ inexact = 1;
+ goto truncate;
+ }
++ else if (bb < cc) /* here sh = 0 and low(b)-low(c) < -0.5 ulp,
++ this corresponds to cases 1g and 1f3 */
++ goto sub_one_ulp;
++ /* the only case where we can't conclude is sh=0 and bb=cc,
++ i.e., we have low(b) - low(c) = -0.5 ulp (up to now), thus
++ we don't know if we must truncate or subtract one ulp.
++ Note: for sh=0 we can't have low(b) - low(c) = -0.5 ulp up to
++ now, since low(b) - low(c) > 1/2^sh */
+ }
+ }
+- else if (bb > cc)
++ else if (cmp_low > 0) /* 0 < low(b) - low(c): either truncate or
++ add one ulp */
+ {
+ if (rnd_mode == MPFR_RNDZ)
+ {
+@@ -437,34 +500,70 @@
+ goto add_one_ulp;
+ else /* round to nearest */
+ {
+- if (is_exact)
++ if (bb > cc)
+ {
+- inexact = -1;
+- goto truncate;
++ /* if sh=0, then bb>cc means that low(b)-low(c) > 0.5 ulp,
++ and similarly when cmp_low=2 */
++ if (cmp_low == 2) /* cases 1a, 1b1, 2a and 2b1 */
++ goto add_one_ulp;
++ /* sh > 0 and cmp_low > 0: this implies that the sh initial
++ neglected bits were 0, and the remaining low(b)-low(c)>0,
++ but its weight is less than 0.5 ulp */
++ else /* 0 < low(b) - low(c) < 0.5 ulp, this corresponds to
++ cases 3a, 1d1 and 3b1 */
++ {
++ inexact = -1;
++ goto truncate;
++ }
+ }
+- else if (down)
++ else if (bb < cc) /* 0 < low(b) - low(c) < 0.5 ulp, cases 1c,
++ 1b3, 2b3 and 2c */
+ {
+- inexact = 1;
++ inexact = -1;
+ goto truncate;
+ }
+- else
+- goto add_one_ulp;
++ /* the only case where we can't conclude is bb=cc, i.e.,
++ low(b) - low(c) = 0.5 ulp (up to now), thus we don't know
++ if we must truncate or add one ulp. */
+ }
+ }
++ /* after k=0, we cannot conclude in the following cases, we split them
++ according to the values of bb and cc for k=1:
++ 1b. sh=0 and cmp_low = 1 and bb-cc = half [around 0.5 ulp]
++ 1b1. bb > cc: add one ulp, inex = 1
++ 1b2: bb = cc: cannot conclude
++ 1b3: bb < cc: truncate, inex = -1
++ 1d. sh=0 and cmp_low = 0 and bb-cc = 0 [around 0]
++ 1d1: bb > cc: truncate, inex = -1
++ 1d2: bb = cc: cannot conclude
++ 1d3: bb < cc: truncate, inex = +1
++ 1f. sh=0 and cmp_low = -1 and bb-cc = -half [around -0.5 ulp]
++ 1f1: bb > cc: truncate, inex = +1
++ 1f2: bb = cc: cannot conclude
++ 1f3: bb < cc: sub one ulp, inex = -1
++ 2b. sh > 0 and cmp_low = 2 and bb=cc [around 0.5 ulp]
++ 2b1. bb > cc: add one ulp, inex = 1
++ 2b2: bb = cc: cannot conclude
++ 2b3: bb < cc: truncate, inex = -1
++ 3b. sh > 0 and cmp_low = 0 [around 0]
++ 3b1. bb > cc: truncate, inex = -1
++ 3b2: bb = cc: cannot conclude
++ 3b3: bb < cc: truncate, inex = +1
++ */
+ }
+
+- if ((rnd_mode == MPFR_RNDN) && !is_exact)
++ if ((rnd_mode == MPFR_RNDN) && cmp_low != 0)
+ {
+ /* even rounding rule */
+ if ((ap[0] >> sh) & 1)
+ {
+- if (down)
++ if (cmp_low < 0)
+ goto sub_one_ulp;
+ else
+ goto add_one_ulp;
+ }
+ else
+- inexact = (down) ? 1 : -1;
++ inexact = (cmp_low > 0) ? -1 : 1;
+ }
+ else
+ inexact = 0;
+diff -Naurd mpfr-3.0.0-a/tests/tfma.c mpfr-3.0.0-b/tests/tfma.c
+--- mpfr-3.0.0-a/tests/tfma.c 2010-06-10 11:00:13.000000000 +0000
++++ mpfr-3.0.0-b/tests/tfma.c 2010-10-21 20:59:32.000000000 +0000
+@@ -337,6 +337,94 @@
+ mpfr_clears (x, y, z, r, (mpfr_ptr) 0);
+ }
+
++static void
++bug20101018 (void)
++{
++ mpfr_t x, y, z, t, u;
++ int i;
++
++ mpfr_init2 (x, 64);
++ mpfr_init2 (y, 64);
++ mpfr_init2 (z, 64);
++ mpfr_init2 (t, 64);
++ mpfr_init2 (u, 64);
++
++ mpfr_set_str (x, "0xf.fffffffffffffffp-14766", 16, MPFR_RNDN);
++ mpfr_set_str (y, "-0xf.fffffffffffffffp+317", 16, MPFR_RNDN);
++ mpfr_set_str (z, "0x8.3ffffffffffe3ffp-14443", 16, MPFR_RNDN);
++ mpfr_set_str (t, "0x8.7ffffffffffc7ffp-14444", 16, MPFR_RNDN);
++ i = mpfr_fma (u, x, y, z, MPFR_RNDN);
++ if (mpfr_cmp (u, t) != 0)
++ {
++ printf ("Wrong result in bug20101018 (a)\n");
++ printf ("Expected ");
++ mpfr_out_str (stdout, 16, 0, t, MPFR_RNDN);
++ printf ("\nGot ");
++ mpfr_out_str (stdout, 16, 0, u, MPFR_RNDN);
++ printf ("\n");
++ exit (1);
++ }
++ if (i <= 0)
++ {
++ printf ("Wrong ternary value in bug20101018 (a)\n");
++ printf ("Expected > 0\n");
++ printf ("Got %d\n", i);
++ exit (1);
++ }
++
++ mpfr_set_str (x, "-0xf.fffffffffffffffp-11420", 16, MPFR_RNDN);
++ mpfr_set_str (y, "0xf.fffffffffffffffp+9863", 16, MPFR_RNDN);
++ mpfr_set_str (z, "0x8.fffff80ffffffffp-1551", 16, MPFR_RNDN);
++ mpfr_set_str (t, "0x9.fffff01ffffffffp-1552", 16, MPFR_RNDN);
++ i = mpfr_fma (u, x, y, z, MPFR_RNDN);
++ if (mpfr_cmp (u, t) != 0)
++ {
++ printf ("Wrong result in bug20101018 (b)\n");
++ printf ("Expected ");
++ mpfr_out_str (stdout, 16, 0, t, MPFR_RNDN);
++ printf ("\nGot ");
++ mpfr_out_str (stdout, 16, 0, u, MPFR_RNDN);
++ printf ("\n");
++ exit (1);
++ }
++ if (i <= 0)
++ {
++ printf ("Wrong ternary value in bug20101018 (b)\n");
++ printf ("Expected > 0\n");
++ printf ("Got %d\n", i);
++ exit (1);
++ }
++
++ mpfr_set_str (x, "0xf.fffffffffffffffp-2125", 16, MPFR_RNDN);
++ mpfr_set_str (y, "-0xf.fffffffffffffffp-6000", 16, MPFR_RNDN);
++ mpfr_set_str (z, "0x8p-8119", 16, MPFR_RNDN);
++ mpfr_set_str (t, "0x8.000000000000001p-8120", 16, MPFR_RNDN);
++ i = mpfr_fma (u, x, y, z, MPFR_RNDN);
++ if (mpfr_cmp (u, t) != 0)
++ {
++ printf ("Wrong result in bug20101018 (c)\n");
++ printf ("Expected ");
++ mpfr_out_str (stdout, 16, 0, t, MPFR_RNDN);
++ printf ("\nGot ");
++ mpfr_out_str (stdout, 16, 0, u, MPFR_RNDN);
++ printf ("\n");
++ exit (1);
++ }
++ if (i <= 0)
++ {
++ printf ("Wrong ternary value in bug20101018 (c)\n");
++ printf ("Expected > 0\n");
++ printf ("Got %d\n", i);
++ exit (1);
++ }
++
++ mpfr_clear (x);
++ mpfr_clear (y);
++ mpfr_clear (z);
++ mpfr_clear (t);
++ mpfr_clear (u);
++}
++
+ int
+ main (int argc, char *argv[])
+ {
+@@ -345,6 +433,8 @@
+
+ tests_start_mpfr ();
+
++ bug20101018 ();
++
+ mpfr_init (x);
+ mpfr_init (s);
+ mpfr_init (y);
+diff -Naurd mpfr-3.0.0-a/tests/tsub.c mpfr-3.0.0-b/tests/tsub.c
+--- mpfr-3.0.0-a/tests/tsub.c 2010-06-10 11:00:13.000000000 +0000
++++ mpfr-3.0.0-b/tests/tsub.c 2010-10-21 20:59:32.000000000 +0000
+@@ -201,6 +201,8 @@
+ if (mpfr_cmp (z, x))
+ {
+ printf ("Error in mpfr_sub (2)\n");
++ printf ("Expected "); mpfr_print_binary (x); puts ("");
++ printf ("Got "); mpfr_print_binary (z); puts ("");
+ exit (1);
+ }
+ mpfr_set_str_binary (x, "1.1110111011110001110111011111111111101000011001011100101100101101");
+@@ -478,6 +480,156 @@
+ mpfr_clear (u);
+ }
+
++/* Bug found by Jakub Jelinek
++ * http://bugzilla.redhat.com/643657
++ * https://gforge.inria.fr/tracker/index.php?func=detail&aid=11301
++ * The consequence can be either an assertion failure (i = 2 in the
++ * testcase below, in debug mode) or an incorrectly rounded value.
++ */
++static void
++bug20101017 (void)
++{
++ mpfr_t a, b, c;
++ int inex;
++ int i;
++
++ mpfr_init2 (a, GMP_NUMB_BITS * 2);
++ mpfr_init2 (b, GMP_NUMB_BITS);
++ mpfr_init2 (c, GMP_NUMB_BITS);
++
++ /* a = 2^(2N) + k.2^(2N-1) + 2^N and b = 1
++ with N = GMP_NUMB_BITS and k = 0 or 1.
++ c = a - b should round to the same value as a. */
++
++ for (i = 2; i <= 3; i++)
++ {
++ mpfr_set_ui_2exp (a, i, GMP_NUMB_BITS - 1, MPFR_RNDN);
++ mpfr_add_ui (a, a, 1, MPFR_RNDN);
++ mpfr_mul_2ui (a, a, GMP_NUMB_BITS, MPFR_RNDN);
++ mpfr_set_ui (b, 1, MPFR_RNDN);
++ inex = mpfr_sub (c, a, b, MPFR_RNDN);
++ mpfr_set (b, a, MPFR_RNDN);
++ if (! mpfr_equal_p (c, b))
++ {
++ printf ("Error in bug20101017 for i = %d.\n", i);
++ printf ("Expected ");
++ mpfr_out_str (stdout, 16, 0, b, MPFR_RNDN);
++ putchar ('\n');
++ printf ("Got ");
++ mpfr_out_str (stdout, 16, 0, c, MPFR_RNDN);
++ putchar ('\n');
++ exit (1);
++ }
++ if (inex >= 0)
++ {
++ printf ("Error in bug20101017 for i = %d: bad inex value.\n", i);
++ printf ("Expected negative, got %d.\n", inex);
++ exit (1);
++ }
++ }
++
++ mpfr_set_prec (a, 64);
++ mpfr_set_prec (b, 129);
++ mpfr_set_prec (c, 2);
++ mpfr_set_str_binary (b, "0.100000000000000000000000000000000000000000000000000000000000000010000000000000000000000000000000000000000000000000000000000000001E65");
++ mpfr_set_str_binary (c, "0.10E1");
++ inex = mpfr_sub (a, b, c, MPFR_RNDN);
++ if (mpfr_cmp_ui_2exp (a, 1, 64) != 0 || inex >= 0)
++ {
++ printf ("Error in mpfr_sub for b-c for b=2^64+1+2^(-64), c=1\n");
++ printf ("Expected result 2^64 with inex < 0\n");
++ printf ("Got "); mpfr_print_binary (a);
++ printf (" with inex=%d\n", inex);
++ exit (1);
++ }
++
++ mpfr_clears (a, b, c, (mpfr_ptr) 0);
++}
++
++/* hard test of rounding */
++static void
++check_rounding (void)
++{
++ mpfr_t a, b, c, res;
++ mpfr_prec_t p;
++ long k, l;
++ int i;
++
++#define MAXKL (2 * GMP_NUMB_BITS)
++ for (p = MPFR_PREC_MIN; p <= GMP_NUMB_BITS; p++)
++ {
++ mpfr_init2 (a, p);
++ mpfr_init2 (res, p);
++ mpfr_init2 (b, p + 1 + MAXKL);
++ mpfr_init2 (c, MPFR_PREC_MIN);
++
++ /* b = 2^p + 1 + 2^(-k), c = 2^(-l) */
++ for (k = 0; k <= MAXKL; k++)
++ for (l = 0; l <= MAXKL; l++)
++ {
++ mpfr_set_ui_2exp (b, 1, p, MPFR_RNDN);
++ mpfr_add_ui (b, b, 1, MPFR_RNDN);
++ mpfr_mul_2ui (b, b, k, MPFR_RNDN);
++ mpfr_add_ui (b, b, 1, MPFR_RNDN);
++ mpfr_div_2ui (b, b, k, MPFR_RNDN);
++ mpfr_set_ui_2exp (c, 1, -l, MPFR_RNDN);
++ i = mpfr_sub (a, b, c, MPFR_RNDN);
++ /* b - c = 2^p + 1 + 2^(-k) - 2^(-l), should be rounded to
++ 2^p for l <= k, and 2^p+2 for l < k */
++ if (l <= k)
++ {
++ if (mpfr_cmp_ui_2exp (a, 1, p) != 0)
++ {
++ printf ("Wrong result in check_rounding\n");
++ printf ("p=%lu k=%ld l=%ld\n", p, k, l);
++ printf ("b="); mpfr_print_binary (b); puts ("");
++ printf ("c="); mpfr_print_binary (c); puts ("");
++ printf ("Expected 2^%lu\n", p);
++ printf ("Got "); mpfr_print_binary (a); puts ("");
++ exit (1);
++ }
++ if (i >= 0)
++ {
++ printf ("Wrong ternary value in check_rounding\n");
++ printf ("p=%lu k=%ld l=%ld\n", p, k, l);
++ printf ("b="); mpfr_print_binary (b); puts ("");
++ printf ("c="); mpfr_print_binary (c); puts ("");
++ printf ("a="); mpfr_print_binary (a); puts ("");
++ printf ("Expected < 0, got %d\n", i);
++ exit (1);
++ }
++ }
++ else /* l < k */
++ {
++ mpfr_set_ui_2exp (res, 1, p, MPFR_RNDN);
++ mpfr_add_ui (res, res, 2, MPFR_RNDN);
++ if (mpfr_cmp (a, res) != 0)
++ {
++ printf ("Wrong result in check_rounding\n");
++ printf ("b="); mpfr_print_binary (b); puts ("");
++ printf ("c="); mpfr_print_binary (c); puts ("");
++ printf ("Expected "); mpfr_print_binary (res); puts ("");
++ printf ("Got "); mpfr_print_binary (a); puts ("");
++ exit (1);
++ }
++ if (i <= 0)
++ {
++ printf ("Wrong ternary value in check_rounding\n");
++ printf ("b="); mpfr_print_binary (b); puts ("");
++ printf ("c="); mpfr_print_binary (c); puts ("");
++ printf ("Expected > 0, got %d\n", i);
++ exit (1);
++ }
++ }
++ }
++
++ mpfr_clear (a);
++ mpfr_clear (res);
++ mpfr_clear (b);
++ mpfr_clear (c);
++ }
++}
++
+ #define TEST_FUNCTION test_sub
+ #define TWO_ARGS
+ #define RAND_FUNCTION(x) mpfr_random2(x, MPFR_LIMB_SIZE (x), randlimb () % 100, RANDS)
+@@ -491,6 +643,8 @@
+
+ tests_start_mpfr ();
+
++ bug20101017 ();
++ check_rounding ();
+ check_diverse ();
+ check_inexact ();
+ bug_ddefour ();
+diff -Naurd mpfr-3.0.0-a/version.c mpfr-3.0.0-b/version.c
+--- mpfr-3.0.0-a/version.c 2010-10-21 20:28:38.000000000 +0000
++++ mpfr-3.0.0-b/version.c 2010-10-21 20:59:32.000000000 +0000
+@@ -25,5 +25,5 @@
+ const char *
+ mpfr_get_version (void)
+ {
+- return "3.0.0-p5";
++ return "3.0.0-p6";
+ }
+diff -Naurd mpfr-3.0.0-a/PATCHES mpfr-3.0.0-b/PATCHES
+--- mpfr-3.0.0-a/PATCHES 2010-10-21 21:18:26.000000000 +0000
++++ mpfr-3.0.0-b/PATCHES 2010-10-21 21:18:26.000000000 +0000
+@@ -0,0 +1 @@
++mpfr_set_ld
+diff -Naurd mpfr-3.0.0-a/VERSION mpfr-3.0.0-b/VERSION
+--- mpfr-3.0.0-a/VERSION 2010-10-21 20:59:32.000000000 +0000
++++ mpfr-3.0.0-b/VERSION 2010-10-21 21:18:26.000000000 +0000
+@@ -1 +1 @@
+-3.0.0-p6
++3.0.0-p7
+diff -Naurd mpfr-3.0.0-a/mpfr.h mpfr-3.0.0-b/mpfr.h
+--- mpfr-3.0.0-a/mpfr.h 2010-10-21 20:59:32.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.h 2010-10-21 21:18:26.000000000 +0000
+@@ -27,7 +27,7 @@
+ #define MPFR_VERSION_MAJOR 3
+ #define MPFR_VERSION_MINOR 0
+ #define MPFR_VERSION_PATCHLEVEL 0
+-#define MPFR_VERSION_STRING "3.0.0-p6"
++#define MPFR_VERSION_STRING "3.0.0-p7"
+
+ /* Macros dealing with MPFR VERSION */
+ #define MPFR_VERSION_NUM(a,b,c) (((a) << 16L) | ((b) << 8) | (c))
+diff -Naurd mpfr-3.0.0-a/set_ld.c mpfr-3.0.0-b/set_ld.c
+--- mpfr-3.0.0-a/set_ld.c 2010-06-10 11:00:14.000000000 +0000
++++ mpfr-3.0.0-b/set_ld.c 2010-10-21 21:18:26.000000000 +0000
+@@ -102,21 +102,25 @@
+ {
+ x /= div13; /* exact */
+ shift_exp += 8192;
++ mpfr_div_2si (t, t, 8192, MPFR_RNDZ);
+ }
+ if (ABS (x) >= div12)
+ {
+ x /= div12; /* exact */
+ shift_exp += 4096;
++ mpfr_div_2si (t, t, 4096, MPFR_RNDZ);
+ }
+ if (ABS (x) >= div11)
+ {
+ x /= div11; /* exact */
+ shift_exp += 2048;
++ mpfr_div_2si (t, t, 2048, MPFR_RNDZ);
+ }
+ if (ABS (x) >= div10)
+ {
+ x /= div10; /* exact */
+ shift_exp += 1024;
++ mpfr_div_2si (t, t, 1024, MPFR_RNDZ);
+ }
+ /* warning: we may have DBL_MAX=2^1024*(1-2^(-53)) < x < 2^1024,
+ therefore we have one extra exponent reduction step */
+@@ -124,9 +128,10 @@
+ {
+ x /= div9; /* exact */
+ shift_exp += 512;
++ mpfr_div_2si (t, t, 512, MPFR_RNDZ);
+ }
+ } /* Check overflow of double */
+- else
++ else /* no overflow on double */
+ {
+ long double div9, div10, div11;
+
+@@ -149,29 +154,34 @@
+ {
+ x /= div13; /* exact */
+ shift_exp -= 8192;
++ mpfr_mul_2si (t, t, 8192, MPFR_RNDZ);
+ }
+ if (ABS (x) <= div12)
+ {
+ x /= div12; /* exact */
+ shift_exp -= 4096;
++ mpfr_mul_2si (t, t, 4096, MPFR_RNDZ);
+ }
+ if (ABS (x) <= div11)
+ {
+ x /= div11; /* exact */
+ shift_exp -= 2048;
++ mpfr_mul_2si (t, t, 2048, MPFR_RNDZ);
+ }
+ if (ABS (x) <= div10)
+ {
+ x /= div10; /* exact */
+ shift_exp -= 1024;
++ mpfr_mul_2si (t, t, 1024, MPFR_RNDZ);
+ }
+ if (ABS(x) <= div9)
+ {
+ x /= div9; /* exact */
+ shift_exp -= 512;
++ mpfr_mul_2si (t, t, 512, MPFR_RNDZ);
+ }
+ }
+- else
++ else /* no underflow */
+ {
+ inexact = mpfr_set_d (u, (double) x, MPFR_RNDZ);
+ MPFR_ASSERTD (inexact == 0);
+diff -Naurd mpfr-3.0.0-a/tests/tset_ld.c mpfr-3.0.0-b/tests/tset_ld.c
+--- mpfr-3.0.0-a/tests/tset_ld.c 2010-06-10 11:00:13.000000000 +0000
++++ mpfr-3.0.0-b/tests/tset_ld.c 2010-10-21 21:18:26.000000000 +0000
+@@ -147,12 +147,39 @@
+ test_fixed_bugs (void)
+ {
+ mpfr_t x;
+- long double d;
++ long double l, m;
+
+ /* bug found by Steve Kargl (2009-03-14) */
+ mpfr_init2 (x, 64);
+ mpfr_set_ui_2exp (x, 1, -16447, MPFR_RNDN);
+- d = mpfr_get_ld (x, MPFR_RNDN); /* an assertion failed in init2.c:50 */
++ mpfr_get_ld (x, MPFR_RNDN); /* an assertion failed in init2.c:50 */
++
++ /* bug reported by Jakub Jelinek (2010-10-17)
++ https://gforge.inria.fr/tracker/?func=detail&aid=11300 */
++ mpfr_set_prec (x, MPFR_LDBL_MANT_DIG);
++ /* l = 0x1.23456789abcdef0123456789abcdp-914L; */
++ l = 8.215640181713713164092636634579e-276;
++ mpfr_set_ld (x, l, MPFR_RNDN);
++ m = mpfr_get_ld (x, MPFR_RNDN);
++ if (m != l)
++ {
++ printf ("Error in get_ld o set_ld for l=%Le\n", l);
++ printf ("Got m=%Le instead of l\n", m);
++ exit (1);
++ }
++
++ /* another similar test which failed with extended double precision and the
++ generic code for mpfr_set_ld */
++ /* l = 0x1.23456789abcdef0123456789abcdp-968L; */
++ l = 4.560596445887084662336528403703e-292;
++ mpfr_set_ld (x, l, MPFR_RNDN);
++ m = mpfr_get_ld (x, MPFR_RNDN);
++ if (m != l)
++ {
++ printf ("Error in get_ld o set_ld for l=%Le\n", l);
++ printf ("Got m=%Le instead of l\n", m);
++ exit (1);
++ }
+
+ mpfr_clear (x);
+ }
+diff -Naurd mpfr-3.0.0-a/version.c mpfr-3.0.0-b/version.c
+--- mpfr-3.0.0-a/version.c 2010-10-21 20:59:32.000000000 +0000
++++ mpfr-3.0.0-b/version.c 2010-10-21 21:18:26.000000000 +0000
+@@ -25,5 +25,5 @@
+ const char *
+ mpfr_get_version (void)
+ {
+- return "3.0.0-p6";
++ return "3.0.0-p7";
+ }
+diff -Naurd mpfr-3.0.0-a/PATCHES mpfr-3.0.0-b/PATCHES
+--- mpfr-3.0.0-a/PATCHES 2010-11-09 15:15:07.000000000 +0000
++++ mpfr-3.0.0-b/PATCHES 2010-11-09 15:15:07.000000000 +0000
+@@ -0,0 +1 @@
++macros
+diff -Naurd mpfr-3.0.0-a/VERSION mpfr-3.0.0-b/VERSION
+--- mpfr-3.0.0-a/VERSION 2010-10-21 21:18:26.000000000 +0000
++++ mpfr-3.0.0-b/VERSION 2010-11-09 15:15:07.000000000 +0000
+@@ -1 +1 @@
+-3.0.0-p7
++3.0.0-p8
+diff -Naurd mpfr-3.0.0-a/mpfr.h mpfr-3.0.0-b/mpfr.h
+--- mpfr-3.0.0-a/mpfr.h 2010-10-21 21:18:26.000000000 +0000
++++ mpfr-3.0.0-b/mpfr.h 2010-11-09 15:15:07.000000000 +0000
+@@ -27,7 +27,7 @@
+ #define MPFR_VERSION_MAJOR 3
+ #define MPFR_VERSION_MINOR 0
+ #define MPFR_VERSION_PATCHLEVEL 0
+-#define MPFR_VERSION_STRING "3.0.0-p7"
++#define MPFR_VERSION_STRING "3.0.0-p8"
+
+ /* Macros dealing with MPFR VERSION */
+ #define MPFR_VERSION_NUM(a,b,c) (((a) << 16L) | ((b) << 8) | (c))
+@@ -67,6 +67,16 @@
+ # define _MPFR_H_HAVE_INTMAX_T 1
+ #endif
+
++/* Avoid some problems with macro expansion if the user defines macros
++ with the same name as keywords. By convention, identifiers and macro
++ names starting with mpfr_ are reserved by MPFR. */
++typedef void mpfr_void;
++typedef int mpfr_int;
++typedef unsigned int mpfr_uint;
++typedef long mpfr_long;
++typedef unsigned long mpfr_ulong;
++typedef size_t mpfr_size_t;
++
+ /* Definition of rounding modes (DON'T USE MPFR_RNDNA!).
+ Warning! Changing the contents of this enum should be seen as an
+ interface change since the old and the new types are not compatible
+@@ -136,7 +146,7 @@
+ typedef mp_exp_t mpfr_exp_t;
+
+ /* Definition of the standard exponent limits */
+-#define MPFR_EMAX_DEFAULT ((mpfr_exp_t) (((unsigned long) 1 << 30) - 1))
++#define MPFR_EMAX_DEFAULT ((mpfr_exp_t) (((mpfr_ulong) 1 << 30) - 1))
+ #define MPFR_EMIN_DEFAULT (-(MPFR_EMAX_DEFAULT))
+
+ /* Definition of the main structure */
+@@ -725,13 +735,13 @@
+ unexpected results with future compilers and aggressive optimisations.
+ Why not working only with signed types, using INT_MIN and LONG_MIN? */
+ #if __GMP_MP_SIZE_T_INT
+-#define __MPFR_EXP_NAN ((mpfr_exp_t)((~((~(unsigned int)0)>>1))+2))
+-#define __MPFR_EXP_ZERO ((mpfr_exp_t)((~((~(unsigned int)0)>>1))+1))
+-#define __MPFR_EXP_INF ((mpfr_exp_t)((~((~(unsigned int)0)>>1))+3))
++#define __MPFR_EXP_NAN ((mpfr_exp_t)((~((~(mpfr_uint)0)>>1))+2))
++#define __MPFR_EXP_ZERO ((mpfr_exp_t)((~((~(mpfr_uint)0)>>1))+1))
++#define __MPFR_EXP_INF ((mpfr_exp_t)((~((~(mpfr_uint)0)>>1))+3))
+ #else
+-#define __MPFR_EXP_NAN ((mpfr_exp_t)((~((~(unsigned long)0)>>1))+2))
+-#define __MPFR_EXP_ZERO ((mpfr_exp_t)((~((~(unsigned long)0)>>1))+1))
+-#define __MPFR_EXP_INF ((mpfr_exp_t)((~((~(unsigned long)0)>>1))+3))
++#define __MPFR_EXP_NAN ((mpfr_exp_t)((~((~(mpfr_ulong)0)>>1))+2))
++#define __MPFR_EXP_ZERO ((mpfr_exp_t)((~((~(mpfr_ulong)0)>>1))+1))
++#define __MPFR_EXP_INF ((mpfr_exp_t)((~((~(mpfr_ulong)0)>>1))+3))
+ #endif
+
+ /* Define MPFR_USE_EXTENSION to avoid "gcc -pedantic" warnings. */
+@@ -760,9 +770,9 @@
+ #define mpfr_inf_p(_x) ((_x)->_mpfr_exp == __MPFR_EXP_INF)
+ #define mpfr_zero_p(_x) ((_x)->_mpfr_exp == __MPFR_EXP_ZERO)
+ #define mpfr_regular_p(_x) ((_x)->_mpfr_exp > __MPFR_EXP_INF)
+-#define mpfr_sgn(_x) \
+- ((_x)->_mpfr_exp < __MPFR_EXP_INF ? \
+- (mpfr_nan_p (_x) ? mpfr_set_erangeflag () : (void) 0), 0 : \
++#define mpfr_sgn(_x) \
++ ((_x)->_mpfr_exp < __MPFR_EXP_INF ? \
++ (mpfr_nan_p (_x) ? mpfr_set_erangeflag () : (mpfr_void) 0), 0 : \
+ MPFR_SIGN (_x))
+
+ /* Prevent them from using as lvalues */
+@@ -805,7 +815,19 @@
+ Moreover casts to unsigned long have been added to avoid warnings in
+ programs that use MPFR and are compiled with -Wconversion; such casts
+ are OK since if X is a constant expression, then (unsigned long) X is
+- also a constant expression, so that the optimizations still work. */
++ also a constant expression, so that the optimizations still work. The
++ warnings are probably related to the following two bugs:
++ http://gcc.gnu.org/bugzilla/show_bug.cgi?id=4210
++ http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38470 (possibly a variant)
++ and the casts could be removed once these bugs are fixed.
++ Casts shouldn't be used on the generic calls (to the ..._2exp functions),
++ where implicit conversions are performed. Indeed, having@least one
++ implicit conversion in the macro allows the compiler to emit diagnostics
++ when normally expected, for instance in the following call:
++ mpfr_set_ui (x, "foo", MPFR_RNDN);
++ If this is not possible (for future macros), one of the tricks described
++ on http://groups.google.com/group/comp.std.c/msg/e92abd24bf9eaf7b could
++ be used. */
+ #if defined (__GNUC__) && !defined(__ICC) && !defined(__cplusplus)
+ #if (__GNUC__ >= 2)
+ #undef mpfr_cmp_ui
+@@ -813,45 +835,45 @@
+ But warning! mpfr_sgn is specified as a macro in the API, thus the macro
+ mustn't be used if side effects are possible, like here. */
+ #define mpfr_cmp_ui(_f,_u) \
+- (__builtin_constant_p (_u) && (unsigned long) (_u) == 0 ? \
++ (__builtin_constant_p (_u) && (mpfr_ulong) (_u) == 0 ? \
+ (mpfr_sgn) (_f) : \
+- mpfr_cmp_ui_2exp ((_f), (unsigned long) (_u), 0))
++ mpfr_cmp_ui_2exp ((_f), (_u), 0))
+ #undef mpfr_cmp_si
+-#define mpfr_cmp_si(_f,_s) \
+- (__builtin_constant_p (_s) && (long) (_s) >= 0 ? \
+- mpfr_cmp_ui ((_f), (unsigned long) (long) (_s)) : \
+- mpfr_cmp_si_2exp ((_f), (long) (_s), 0))
++#define mpfr_cmp_si(_f,_s) \
++ (__builtin_constant_p (_s) && (mpfr_long) (_s) >= 0 ? \
++ mpfr_cmp_ui ((_f), (mpfr_ulong) (mpfr_long) (_s)) : \
++ mpfr_cmp_si_2exp ((_f), (_s), 0))
+ #if __GNUC__ > 2 || __GNUC_MINOR__ >= 95
+ #undef mpfr_set_ui
+ #define mpfr_set_ui(_f,_u,_r) \
+- (__builtin_constant_p (_u) && (unsigned long) (_u) == 0 ? \
++ (__builtin_constant_p (_u) && (mpfr_ulong) (_u) == 0 ? \
+ __extension__ ({ \
+ mpfr_ptr _p = (_f); \
+ _p->_mpfr_sign = 1; \
+ _p->_mpfr_exp = __MPFR_EXP_ZERO; \
+- (void) (_r); 0; }) : \
+- mpfr_set_ui_2exp ((_f), (unsigned long) (_u), 0, (_r)))
++ (mpfr_void) (_r); 0; }) : \
++ mpfr_set_ui_2exp ((_f), (_u), 0, (_r)))
+ #endif
+ #undef mpfr_set_si
+ #define mpfr_set_si(_f,_s,_r) \
+- (__builtin_constant_p (_s) && (long) (_s) >= 0 ? \
+- mpfr_set_ui ((_f), (unsigned long) (long) (_s), (_r)) : \
+- mpfr_set_si_2exp ((_f), (long) (_s), 0, (_r)))
++ (__builtin_constant_p (_s) && (mpfr_long) (_s) >= 0 ? \
++ mpfr_set_ui ((_f), (mpfr_ulong) (mpfr_long) (_s), (_r)) : \
++ mpfr_set_si_2exp ((_f), (_s), 0, (_r)))
+ #endif
+ #endif
+
+ /* Macro version of mpfr_stack interface for fast access */
+-#define mpfr_custom_get_size(p) ((size_t) \
++#define mpfr_custom_get_size(p) ((mpfr_size_t) \
+ (((p)+GMP_NUMB_BITS-1)/GMP_NUMB_BITS*sizeof (mp_limb_t)))
+ #define mpfr_custom_init(m,p) do {} while (0)
+-#define mpfr_custom_get_significand(x) ((void*)((x)->_mpfr_d))
++#define mpfr_custom_get_significand(x) ((mpfr_void*)((x)->_mpfr_d))
+ #define mpfr_custom_get_exp(x) ((x)->_mpfr_exp)
+ #define mpfr_custom_move(x,m) do { ((x)->_mpfr_d = (mp_limb_t*)(m)); } while (0)
+ #define mpfr_custom_init_set(x,k,e,p,m) do { \
+ mpfr_ptr _x = (x); \
+ mpfr_exp_t _e; \
+ mpfr_kind_t _t; \
+- int _s, _k; \
++ mpfr_int _s, _k; \
+ _k = (k); \
+ if (_k >= 0) { \
+ _t = (mpfr_kind_t) _k; \
+@@ -868,11 +890,13 @@
+ _x->_mpfr_exp = _e; \
+ _x->_mpfr_d = (mp_limb_t*) (m); \
+ } while (0)
+-#define mpfr_custom_get_kind(x) \
+- ( (x)->_mpfr_exp > __MPFR_EXP_INF ? (int)MPFR_REGULAR_KIND*MPFR_SIGN (x) \
+- : (x)->_mpfr_exp == __MPFR_EXP_INF ? (int)MPFR_INF_KIND*MPFR_SIGN (x) \
+- : (x)->_mpfr_exp == __MPFR_EXP_NAN ? (int)MPFR_NAN_KIND \
+- : (int) MPFR_ZERO_KIND * MPFR_SIGN (x) )
++#define mpfr_custom_get_kind(x) \
++ ( (x)->_mpfr_exp > __MPFR_EXP_INF ? \
++ (mpfr_int) MPFR_REGULAR_KIND * MPFR_SIGN (x) \
++ : (x)->_mpfr_exp == __MPFR_EXP_INF ? \
++ (mpfr_int) MPFR_INF_KIND * MPFR_SIGN (x) \
++ : (x)->_mpfr_exp == __MPFR_EXP_NAN ? (mpfr_int) MPFR_NAN_KIND \
++ : (mpfr_int) MPFR_ZERO_KIND * MPFR_SIGN (x) )
+
+
+ #endif /* MPFR_USE_NO_MACRO */
+diff -Naurd mpfr-3.0.0-a/version.c mpfr-3.0.0-b/version.c
+--- mpfr-3.0.0-a/version.c 2010-10-21 21:18:26.000000000 +0000
++++ mpfr-3.0.0-b/version.c 2010-11-09 15:15:07.000000000 +0000
+@@ -25,5 +25,5 @@
+ const char *
+ mpfr_get_version (void)
+ {
+- return "3.0.0-p7";
++ return "3.0.0-p8";
+ }
diff --git a/package/mpfr/mpfr.mk b/package/mpfr/mpfr.mk
index 6979a9f..15ee48a 100644
--- a/package/mpfr/mpfr.mk
+++ b/package/mpfr/mpfr.mk
@@ -3,122 +3,14 @@
# mpfr
#
#############################################################
-MPFR_VERSION:=2.4.1
-MPFR_PATCH:=patches
-MPFR_PATCH_FILE:=mpfr-$(MPFR_VERSION).patch
-MPFR_SOURCE:=mpfr-$(MPFR_VERSION).tar.bz2
-MPFR_CAT:=$(BZCAT)
-MPFR_SITE:=http://www.mpfr.org/mpfr-$(MPFR_VERSION)
-MPFR_DIR:=$(TOOLCHAIN_DIR)/mpfr-$(MPFR_VERSION)
-MPFR_TARGET_DIR:=$(BUILD_DIR)/mpfr-$(MPFR_VERSION)
-MPFR_BINARY:=libmpfr$(LIBTGTEXT)
-MPFR_HOST_BINARY:=libmpfr$(HOST_LIBEXT)
-MPFR_LIBVERSION:=1.2.0
-# need patch
-ifneq ($(MPFR_PATCH),)
-MPFR_PATCH_SOURCE:=$(DL_DIR)/$(MPFR_PATCH_FILE)
+MPFR_VERSION = 3.0.0
+MPFR_SITE = http://www.mpfr.org/mpfr-$(MPFR_VERSION)
+MPFR_SOURCE = mpfr-$(MPFR_VERSION).tar.bz2
+MPFR_INSTALL_STAGING = YES
+MPFR_DEPENDENCIES = gmp
+MPFR_MAKE_OPT = RANLIB=$(TARGET_RANLIB)
+HOST_MPFR_DEPENDENCIES = host-gmp
-$(MPFR_PATCH_SOURCE):
- $(call DOWNLOAD,$(MPFR_SITE),$(MPFR_PATCH))
-ifeq ($(DL_MODE),DOWNLOAD)
- mv $(DL_DIR)/$(MPFR_PATCH) $@
-endif
-endif
-
-$(DL_DIR)/$(MPFR_SOURCE):
- $(call DOWNLOAD,$(MPFR_SITE),$(MPFR_SOURCE))
-
-$(MPFR_DIR)/.unpacked: $(DL_DIR)/$(MPFR_SOURCE) $(MPFR_PATCH_SOURCE)
- $(MPFR_CAT) $(DL_DIR)/$(MPFR_SOURCE) | tar -C $(TOOLCHAIN_DIR) $(TAR_OPTIONS) -
- toolchain/patch-kernel.sh $(MPFR_DIR) package/mpfr/ \*.patch
- $(CONFIG_UPDATE) $(@D)
-ifneq ($(MPFR_PATCH),)
- toolchain/patch-kernel.sh $(MPFR_DIR) $(DL_DIR)/ $(MPFR_PATCH_FILE)
-endif
- touch $@
-
-$(MPFR_TARGET_DIR)/.configured: $(MPFR_DIR)/.unpacked $(STAGING_DIR)/usr/lib/$(GMP_BINARY)
- mkdir -p $(MPFR_TARGET_DIR)
- (cd $(MPFR_TARGET_DIR); rm -rf config.cache; \
- $(TARGET_CONFIGURE_OPTS) \
- $(TARGET_CONFIGURE_ARGS) \
- $(MPFR_DIR)/configure $(QUIET) \
- --target=$(GNU_TARGET_NAME) \
- --host=$(GNU_TARGET_NAME) \
- --build=$(GNU_HOST_NAME) \
- --prefix=/usr \
- $(PREFERRED_LIB_FLAGS) \
- --with-gmp-build=$(GMP_TARGET_DIR) \
- $(DISABLE_NLS) \
- )
- touch $@
-
-$(MPFR_TARGET_DIR)/.libs/$(MPFR_BINARY): $(MPFR_TARGET_DIR)/.configured
- #$(MAKE) $(TARGET_CONFIGURE_OPTS) -C $(MPFR_TARGET_DIR)
- $(MAKE) -C $(MPFR_TARGET_DIR)
-
-$(STAGING_DIR)/usr/lib/$(MPFR_BINARY): $(MPFR_TARGET_DIR)/.libs/$(MPFR_BINARY)
- $(MAKE) DESTDIR=$(STAGING_DIR) -C $(MPFR_TARGET_DIR) install
- $(STRIPCMD) $(STRIP_STRIP_UNNEEDED) $(STAGING_DIR)/usr/lib/libmpfr$(LIBTGTEXT)*
-
-$(TARGET_DIR)/usr/lib/libmpfr.so $(TARGET_DIR)/usr/lib/libmpfr.so.$(MPFR_LIBVERSION) $(TARGET_DIR)/usr/lib/libmpfr.a: $(STAGING_DIR)/usr/lib/$(MPFR_BINARY)
- cp -dpf $(STAGING_DIR)/usr/lib/libmpfr$(LIBTGTEXT)* $(TARGET_DIR)/usr/lib/
-ifeq ($(BR2_PACKAGE_LIBMPFR_HEADERS),y)
- cp -dpf $(STAGING_DIR)/usr/include/mpfr.h $(STAGING_DIR)/usr/include/mpf2mpfr.h \
- $(TARGET_DIR)/usr/include/
-endif
-
-.PHONY: libmpfr-source
-
-libmpfr-source: $(DL_DIR)/$(MPFR_SOURCE) $(MPFR_PATCH_SOURCE)
-
-libmpfr: $(TARGET_DIR)/usr/lib/libmpfr$(LIBTGTEXT)
-stage-libmpfr: $(STAGING_DIR)/usr/lib/$(MPFR_BINARY)
-
-libmpfr-clean:
- rm -f $(TARGET_DIR)/usr/lib/libmpfr.* \
- $(TARGET_DIR)/usr/include/mpfr.h \
- $(TARGET_DIR)/usr/include/mpf2mpfr.h \
- $(STAGING_DIR)/usr/lib/libmpfr* $(STAGING_DIR)/usr/include/mpfr*
- -$(MAKE) -C $(MPFR_TARGET_DIR) clean
-
-libmpfr-dirclean:
- rm -rf $(MPFR_TARGET_DIR)
-
-MPFR_DIR2:=$(TOOLCHAIN_DIR)/mpfr-$(MPFR_VERSION)-host
-MPFR_HOST_DIR:=$(TOOLCHAIN_DIR)/mpfr
-$(MPFR_DIR2)/.configured: $(MPFR_DIR)/.unpacked $(GMP_HOST_DIR)/lib/$(GMP_HOST_BINARY)
- mkdir -p $(MPFR_DIR2)
- (cd $(MPFR_DIR2); \
- $(HOST_CONFIGURE_OPTS) \
- $(MPFR_DIR)/configure $(QUIET) \
- --prefix="$(MPFR_HOST_DIR)" \
- --build=$(GNU_HOST_NAME) \
- --host=$(GNU_HOST_NAME) \
- --disable-shared \
- --enable-static \
- --with-gmp=$(GMP_HOST_DIR) \
- $(DISABLE_NLS) \
- )
- touch $@
-
-$(MPFR_HOST_DIR)/lib/libmpfr$(HOST_LIBEXT): $(MPFR_DIR2)/.configured
- $(MAKE) -C $(MPFR_DIR2) install
-
-host-libmpfr: $(MPFR_HOST_DIR)/lib/$(MPFR_HOST_BINARY)
-host-libmpfr-source: libmpfr-source
-host-libmpfr-clean:
- rm -rf $(MPFR_HOST_DIR)
- -$(MAKE) -C $(MPFR_DIR2) clean
-host-libmpfr-dirclean:
- rm -rf $(MPFR_HOST_DIR) $(MPFR_DIR2)
-
-#############################################################
-#
-# Toplevel Makefile options
-#
-#############################################################
-ifeq ($(BR2_PACKAGE_LIBMPFR),y)
-TARGETS+=libmpfr
-endif
+$(eval $(call AUTOTARGETS,package,mpfr))
+$(eval $(call AUTOTARGETS,package,mpfr,host))
diff --git a/toolchain/binutils/binutils.mk b/toolchain/binutils/binutils.mk
index 327a0f6..c332590 100644
--- a/toolchain/binutils/binutils.mk
+++ b/toolchain/binutils/binutils.mk
@@ -16,16 +16,16 @@ endif
BINUTILS_HOST_PREREQ:=
BINUTILS_TARGET_PREREQ:=
-BINUTILS_HOST_PREREQ:=host-gmp $(TOOLCHAIN_DIR)/mpfr/lib/libmpfr$(HOST_LIBEXT)
-HOST_SOURCE += host-gmp-source host-libmpfr-source
+BINUTILS_HOST_PREREQ:=host-gmp host-mpfr
+HOST_SOURCE += host-gmp-source host-mpfr-source
-BINUTILS_TARGET_PREREQ:=gmp $(TARGET_DIR)/usr/lib/libmpfr$(LIBTGTEXT)
+BINUTILS_TARGET_PREREQ:=gmp mpfr
EXTRA_BINUTILS_CONFIG_OPTIONS+=--with-gmp=$(HOST_DIR)/usr
-EXTRA_BINUTILS_CONFIG_OPTIONS+=--with-mpfr="$(MPFR_HOST_DIR)"
+EXTRA_BINUTILS_CONFIG_OPTIONS+=--with-mpfr=$(HOST_DIR)/usr
BINUTILS_TARGET_CONFIG_OPTIONS=--with-gmp=$(STAGING_DIR)/usr
-BINUTILS_TARGET_CONFIG_OPTIONS+=--with-mpfr="$(MPFR_TARGET_DIR)"
+BINUTILS_TARGET_CONFIG_OPTIONS+=--with-mpfr=$(STAGING_DIR)/usr
ifeq ($(BR2_PACKAGE_LIBMPC),y)
BINUTILS_ADD_MPC = y
diff --git a/toolchain/gcc/gcc-uclibc-4.x.mk b/toolchain/gcc/gcc-uclibc-4.x.mk
index cbcaadb..ede5b7c 100644
--- a/toolchain/gcc/gcc-uclibc-4.x.mk
+++ b/toolchain/gcc/gcc-uclibc-4.x.mk
@@ -107,8 +107,8 @@ GCC_TARGET_LANGUAGES:=$(GCC_TARGET_LANGUAGES),objc
endif
GCC_WITH_HOST_GMP = --with-gmp=$(HOST_DIR)/usr
-GCC_WITH_HOST_MPFR=--with-mpfr=$(MPFR_HOST_DIR)
-HOST_SOURCE += host-gmp-source host-libmpfr-source
+GCC_WITH_HOST_MPFR = --with-mpfr=$(HOST_DIR)/usr
+HOST_SOURCE += host-gmp-source host-mpfr-source
ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.)
GCC_WITH_HOST_MPC=--with-mpc=$(MPC_HOST_DIR)
@@ -120,7 +120,7 @@ GCC_TARGET_LANGUAGES:=$(GCC_TARGET_LANGUAGES),fortran
#GCC_TARGET_PREREQ+=$(TARGET_DIR)/usr/lib/libmpfr.so $(TARGET_DIR)/usr/lib/libgmp.so
#GCC_STAGING_PREREQ+=$(TOOLCHAIN_DIR)/mpfr/lib/libmpfr.so
GCC_WITH_TARGET_GMP = --with-gmp=$(STAGING_DIR)/usr
-GCC_WITH_TARGET_MPFR=--with-mpfr="$(MPFR_TARGET_DIR)"
+GCC_WITH_TARGET_MPFR = --with-mpfr=$(STAGING_DIR)/usr
ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.)
GCC_WITH_TARGET_MPC=--with-mpc="$(MPC_TARGET_DIR)"
endif
--
1.7.2.2
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [Buildroot] [PATCH 4/5] mpc: make it a proper package
2010-12-06 19:40 [Buildroot] [PATCH 0/5] toolchain rework, part 1 Gustavo Zacarias
` (2 preceding siblings ...)
2010-12-06 19:40 ` [Buildroot] [PATCH 3/5] mpfr: make it a proper package and bump to version 3.0.0-p8 Gustavo Zacarias
@ 2010-12-06 19:40 ` Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 5/5] package infra: fix versioned patches for host packages Gustavo Zacarias
` (2 subsequent siblings)
6 siblings, 0 replies; 8+ messages in thread
From: Gustavo Zacarias @ 2010-12-06 19:40 UTC (permalink / raw)
To: buildroot
* Convert mpc to a proper autotargets package
* Hook it up for binutils/gcc to use
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
---
package/mpc/Config.in | 15 +----
package/mpc/mpc.mk | 109 +++------------------------------------
toolchain/binutils/binutils.mk | 10 ++--
toolchain/gcc/gcc-uclibc-4.x.mk | 6 +-
4 files changed, 18 insertions(+), 122 deletions(-)
diff --git a/package/mpc/Config.in b/package/mpc/Config.in
index 61c9638..8377013 100644
--- a/package/mpc/Config.in
+++ b/package/mpc/Config.in
@@ -1,19 +1,10 @@
-config BR2_PACKAGE_LIBMPC
+config BR2_PACKAGE_MPC
bool "mpc"
- select BR2_PACKAGE_LIBMPFR
- select BR2_PACKAGE_LIBGMP
+ select BR2_PACKAGE_MPFR
+ select BR2_PACKAGE_GMP
help
Mpc is a C library for the arithmetic of complex numbers with
arbitrarily high precision and correct rounding of the result.
It is built upon and follows the same principles as Mpfr.
http://www.multiprecision.org/
-
-
-config BR2_PACKAGE_LIBMPC_HEADERS
- bool "mpc headers for target"
- depends on BR2_PACKAGE_LIBMPC
- select BR2_PACKAGE_LIBMPFR_HEADERS
- select BR2_PACKAGE_LIBGMP_HEADERS
- help
- Install the mpc headers for the target.
diff --git a/package/mpc/mpc.mk b/package/mpc/mpc.mk
index 41c0ffa..8c6d486 100644
--- a/package/mpc/mpc.mk
+++ b/package/mpc/mpc.mk
@@ -3,107 +3,12 @@
# mpc
#
#############################################################
-MPC_VERSION:=0.8.2
-MPC_SOURCE:=mpc-$(MPC_VERSION).tar.gz
-MPC_SITE:=http://www.multiprecision.org/mpc/download
-MPC_CAT:=$(ZCAT)
-MPC_DIR:=$(TOOLCHAIN_DIR)/mpc-$(MPC_VERSION)
-MPC_TARGET_DIR:=$(BUILD_DIR)/mpc-$(MPC_VERSION)
-MPC_BINARY:=libmpc$(LIBTGTEXT)
-MPC_HOST_BINARY:=libmpc$(HOST_LIBEXT)
-MPC_LIBVERSION:=2.0.0
-$(DL_DIR)/$(MPC_SOURCE):
- $(call DOWNLOAD,$(MPC_SITE),$(MPC_SOURCE))
+MPC_VERSION = 0.8.2
+MPC_SITE = http://www.multiprecision.org/mpc/download
+MPC_INSTALL_STAGING = YES
+MPC_DEPENDENCIES = gmp mpfr
+HOST_MPC_DEPENDENCIES = host-gmp host-mpfr
-libmpc-source: $(DL_DIR)/$(MPC_SOURCE)
-
-$(MPC_DIR)/.unpacked: $(DL_DIR)/$(MPC_SOURCE)
- $(MPC_CAT) $(DL_DIR)/$(MPC_SOURCE) | tar -C $(TOOLCHAIN_DIR) $(TAR_OPTIONS) -
- toolchain/patch-kernel.sh $(MPC_DIR) package/mpc/ \*.patch
- $(CONFIG_UPDATE) $(@D)
- touch $@
-
-$(MPC_TARGET_DIR)/.configured: $(MPC_DIR)/.unpacked
- mkdir -p $(MPC_TARGET_DIR)
- (cd $(MPC_TARGET_DIR); rm -rf config.cache; \
- $(TARGET_CONFIGURE_OPTS) \
- $(TARGET_CONFIGURE_ARGS) \
- $(MPC_DIR)/configure $(QUIET) \
- --target=$(GNU_TARGET_NAME) \
- --host=$(GNU_TARGET_NAME) \
- --build=$(GNU_HOST_NAME) \
- --prefix=/usr \
- $(PREFERRED_LIB_FLAGS) \
- --with-mpfr-lib=$(TARGET_DIR)/usr/lib/ \
- --with-gmp-lib=$(TARGET_DIR)/usr/lib/ \
- --with-mpfr-include=$(STAGING_DIR)/usr/include/ \
- --with-gmp-include=$(STAGING_DIR)/usr/include/ \
- $(DISABLE_NLS) \
- )
- touch $@
-
-$(MPC_TARGET_DIR)/src/.libs/$(MPC_BINARY): $(MPC_TARGET_DIR)/.configured
- #$(MAKE) $(TARGET_CONFIGURE_OPTS) -C $(MPC_TARGET_DIR)
- $(MAKE) -C $(MPC_TARGET_DIR)
-
-$(STAGING_DIR)/usr/lib/$(MPC_BINARY): $(MPC_TARGET_DIR)/src/.libs/$(MPC_BINARY)
- $(MAKE) DESTDIR=$(STAGING_DIR) -C $(MPC_TARGET_DIR) install
- $(STRIPCMD) $(STRIP_STRIP_UNNEEDED) $(STAGING_DIR)/usr/lib/libmpc$(LIBTGTEXT)*
-
-$(TARGET_DIR)/usr/lib/libmpc.so $(TARGET_DIR)/usr/lib/libmpc.so.$(MPC_LIBVERSION): $(STAGING_DIR)/usr/lib/$(MPC_BINARY)
- cp -dpf $(STAGING_DIR)/usr/lib/libmpc$(LIBTGTEXT)* $(TARGET_DIR)/usr/lib/
-ifeq ($(BR2_PACKAGE_LIBMPC_HEADERS),y)
- test -d $(TARGET_DIR)/usr/include || mkdir -p $(TARGET_DIR)/usr/include
- cp -dpf $(STAGING_DIR)/usr/include/mpc.h $(TARGET_DIR)/usr/include/
-endif
-
-libmpc: libgmp libmpfr $(TARGET_DIR)/usr/lib/libmpc$(LIBTGTEXT)
-stage-libmpc: $(STAGING_DIR)/usr/lib/$(MPC_BINARY)
-
-libmpc-clean:
- rm -f $(TARGET_DIR)/usr/lib/libmpc.* $(TARGET_DIR)/usr/include/mpc.h \
- $(STAGING_DIR)/usr/lib/libmpc* $(STAGING_DIR)/usr/include/mpc.h
- -$(MAKE) -C $(MPC_TARGET_DIR) clean
-
-libmpc-dirclean:
- rm -rf $(MPC_TARGET_DIR) $(MPC_DIR)
-
-MPC_DIR2:=$(TOOLCHAIN_DIR)/mpc-$(MPC_VERSION)-host
-MPC_HOST_DIR:=$(TOOLCHAIN_DIR)/mpc
-$(MPC_DIR2)/.configured: $(MPC_DIR)/.unpacked
- mkdir -p $(MPC_DIR2)
- (cd $(MPC_DIR2); rm -rf config.cache; \
- $(HOST_CONFIGURE_OPTS) \
- $(MPC_CPP_FLAGS) \
- $(MPC_DIR)/configure $(QUIET) \
- --prefix="$(MPC_HOST_DIR)" \
- --build=$(GNU_HOST_NAME) \
- --host=$(GNU_HOST_NAME) \
- --disable-shared \
- --enable-static \
- --with-mpfr=$(MPFR_HOST_DIR) \
- --with-gmp=$(GMP_HOST_DIR) \
- $(DISABLE_NLS) \
- )
- touch $@
-
-$(MPC_HOST_DIR)/lib/libmpc$(HOST_LIBEXT): $(MPC_DIR2)/.configured
- $(MAKE) -C $(MPC_DIR2) install
-
-host-libmpc: $(MPC_HOST_DIR)/lib/$(MPC_HOST_BINARY)
-host-libmpc-source: libmpc-source
-host-libmpc-clean:
- rm -rf $(MPC_HOST_DIR)
- -$(MAKE) -C $(MPC_DIR2) clean
-host-libmpc-dirclean:
- rm -rf $(MPC_HOST_DIR) $(MPC_DIR2)
-
-#############################################################
-#
-# Toplevel Makefile options
-#
-#############################################################
-ifeq ($(BR2_PACKAGE_LIBMPC),y)
-TARGETS+=libmpc
-endif
+$(eval $(call AUTOTARGETS,package,mpc))
+$(eval $(call AUTOTARGETS,package,mpc,host))
diff --git a/toolchain/binutils/binutils.mk b/toolchain/binutils/binutils.mk
index c332590..8ab383b 100644
--- a/toolchain/binutils/binutils.mk
+++ b/toolchain/binutils/binutils.mk
@@ -36,11 +36,11 @@ BINUTILS_ADD_MPC = y
endif
ifeq ($(BINUTILS_ADD_MPC),y)
-BINUTILS_HOST_PREREQ += $(TOOLCHAIN_DIR)/mpc/lib/libmpc$(HOST_LIBEXT)
-HOST_SOURCE += host-libmpc-source
-BINUTILS_TARGET_PREREQ += $(TARGET_DIR)/usr/lib/libmpc$(LIBTGTEXT)
-EXTRA_BINUTILS_CONFIG_OPTIONS += --with-mpc="$(MPC_HOST_DIR)"
-BINUTILS_TARGET_CONFIG_OPTIONS += --with-mpc="$(MPC_TARGET_DIR)"
+BINUTILS_HOST_PREREQ += host-mpc
+HOST_SOURCE += host-mpc-source
+BINUTILS_TARGET_PREREQ += mpc
+EXTRA_BINUTILS_CONFIG_OPTIONS += --with-mpc=$(HOST_DIR)/usr
+BINUTILS_TARGET_CONFIG_OPTIONS += --with-mpc=$(STAGING_DIR)/usr
endif
BINUTILS_PATCH_DIR:=toolchain/binutils/$(BINUTILS_VERSION)
diff --git a/toolchain/gcc/gcc-uclibc-4.x.mk b/toolchain/gcc/gcc-uclibc-4.x.mk
index ede5b7c..6add764 100644
--- a/toolchain/gcc/gcc-uclibc-4.x.mk
+++ b/toolchain/gcc/gcc-uclibc-4.x.mk
@@ -111,8 +111,8 @@ GCC_WITH_HOST_MPFR = --with-mpfr=$(HOST_DIR)/usr
HOST_SOURCE += host-gmp-source host-mpfr-source
ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.)
-GCC_WITH_HOST_MPC=--with-mpc=$(MPC_HOST_DIR)
-HOST_SOURCE += host-libmpc-source
+GCC_WITH_HOST_MPC = --with-mpc=$(HOST_DIR)/usr
+HOST_SOURCE += host-mpc-source
endif
ifeq ($(BR2_INSTALL_FORTRAN),y)
@@ -122,7 +122,7 @@ GCC_TARGET_LANGUAGES:=$(GCC_TARGET_LANGUAGES),fortran
GCC_WITH_TARGET_GMP = --with-gmp=$(STAGING_DIR)/usr
GCC_WITH_TARGET_MPFR = --with-mpfr=$(STAGING_DIR)/usr
ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.)
-GCC_WITH_TARGET_MPC=--with-mpc="$(MPC_TARGET_DIR)"
+GCC_WITH_TARGET_MPC = --with-mpc=$(STAGING_DIR)/usr
endif
endif
--
1.7.2.2
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [Buildroot] [PATCH 5/5] package infra: fix versioned patches for host packages
2010-12-06 19:40 [Buildroot] [PATCH 0/5] toolchain rework, part 1 Gustavo Zacarias
` (3 preceding siblings ...)
2010-12-06 19:40 ` [Buildroot] [PATCH 4/5] mpc: make it a proper package Gustavo Zacarias
@ 2010-12-06 19:40 ` Gustavo Zacarias
2010-12-06 23:07 ` [Buildroot] [PATCH 0/5] toolchain rework, part 1 Thomas Petazzoni
2010-12-27 22:25 ` Peter Korsgaard
6 siblings, 0 replies; 8+ messages in thread
From: Gustavo Zacarias @ 2010-12-06 19:40 UTC (permalink / raw)
To: buildroot
Without this patch we'd need separate versioned patch directories for
target and host packages, thus having unwanted duplication.
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
---
package/Makefile.package.in | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/package/Makefile.package.in b/package/Makefile.package.in
index 816a013..4cb0334 100644
--- a/package/Makefile.package.in
+++ b/package/Makefile.package.in
@@ -250,7 +250,7 @@ $(BUILD_DIR)/%/.stamp_patched:
toolchain/patch-kernel.sh $(@D) $($(PKG)_DIR_PREFIX)/$($(NOHOSTPKG)_NAME) $(NAMEVER)\*.patch $(NAMEVER)\*.patch.$(ARCH) || exit 1; \
else \
toolchain/patch-kernel.sh $(@D) $($(PKG)_DIR_PREFIX)/$($(NOHOSTPKG)_NAME) $($(NOHOSTPKG)_NAME)\*.patch $($(NOHOSTPKG)_NAME)\*.patch.$(ARCH) || exit 1; \
- if test -d $($(PKG)_DIR_PREFIX)/$($(PKG)_NAME)/$(NAMEVER); then \
+ if test -d $($(PKG)_DIR_PREFIX)/$($(NOHOSTPKG)_NAME)/$(NAMEVER); then \
toolchain/patch-kernel.sh $(@D) $($(PKG)_DIR_PREFIX)/$($(NOHOSTPKG)_NAME)/$(NAMEVER) \*.patch \*.patch.$(ARCH) || exit 1; \
fi; \
fi; \
--
1.7.2.2
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [Buildroot] [PATCH 0/5] toolchain rework, part 1
2010-12-06 19:40 [Buildroot] [PATCH 0/5] toolchain rework, part 1 Gustavo Zacarias
` (4 preceding siblings ...)
2010-12-06 19:40 ` [Buildroot] [PATCH 5/5] package infra: fix versioned patches for host packages Gustavo Zacarias
@ 2010-12-06 23:07 ` Thomas Petazzoni
2010-12-27 22:25 ` Peter Korsgaard
6 siblings, 0 replies; 8+ messages in thread
From: Thomas Petazzoni @ 2010-12-06 23:07 UTC (permalink / raw)
To: buildroot
Hello,
On Mon, 6 Dec 2010 16:40:06 -0300
Gustavo Zacarias <gustavo@zacarias.com.ar> wrote:
> This is the first batch of patches for the internal toolchain rework.
>
> In this first stage we convert gmp, mpfr and mpc to proper packages and
> hook them up for use in the current toolchain.
>
> We also fix a small bug in the package infra with regards to versioned patches
> that'll be required to avoid duplication in the binutils/gcc migration.
I have reviewed these patches, and tested that I could build
successfully an ARM toolchain, and a native ARM toolchain for the
target. The ARM cross-compiling toolchain was used to compile a
complete DirectFB+Gtk stack. The native toolchain on the target was
used to compile a simple hello world program, that I could execute.
Therefore, since I have also reviewed those changes, they all get my:
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Regards,
Thomas
--
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Buildroot] [PATCH 0/5] toolchain rework, part 1
2010-12-06 19:40 [Buildroot] [PATCH 0/5] toolchain rework, part 1 Gustavo Zacarias
` (5 preceding siblings ...)
2010-12-06 23:07 ` [Buildroot] [PATCH 0/5] toolchain rework, part 1 Thomas Petazzoni
@ 2010-12-27 22:25 ` Peter Korsgaard
6 siblings, 0 replies; 8+ messages in thread
From: Peter Korsgaard @ 2010-12-27 22:25 UTC (permalink / raw)
To: buildroot
>>>>> "Gustavo" == Gustavo Zacarias <gustavo@zacarias.com.ar> writes:
Gustavo> This is the first batch of patches for the internal toolchain
Gustavo> rework. In this first stage we convert gmp, mpfr and mpc to
Gustavo> proper packages and hook them up for use in the current
Gustavo> toolchain.
Gustavo> We also fix a small bug in the package infra with regards to
Gustavo> versioned patches that'll be required to avoid duplication in
Gustavo> the binutils/gcc migration.
Committed, thanks.
--
Bye, Peter Korsgaard
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2010-12-27 22:25 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-12-06 19:40 [Buildroot] [PATCH 0/5] toolchain rework, part 1 Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 1/5] gmp: make it a proper package and bump to version 5.0.1 Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 2/5] php: libgmp was renamed to gmp Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 3/5] mpfr: make it a proper package and bump to version 3.0.0-p8 Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 4/5] mpc: make it a proper package Gustavo Zacarias
2010-12-06 19:40 ` [Buildroot] [PATCH 5/5] package infra: fix versioned patches for host packages Gustavo Zacarias
2010-12-06 23:07 ` [Buildroot] [PATCH 0/5] toolchain rework, part 1 Thomas Petazzoni
2010-12-27 22:25 ` Peter Korsgaard
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox