[SOLARIS 11, SPARC] compilation failed #1080

issue tankf33der openend this issue on February 25, 2022

tankf33der commented at 2:45 pm on February 25, 2022: none

  0$ nice ./configure 
  1checking build system type... sparc-sun-solaris2.11
  2checking host system type... sparc-sun-solaris2.11
  3checking for a BSD-compatible install... /usr/bin/ginstall -c
  4checking whether build environment is sane... yes
  5checking for a thread-safe mkdir -p... /usr/bin/gmkdir -p
  6checking for gawk... gawk
  7checking whether make sets $(MAKE)... yes
  8checking whether make supports nested variables... yes
  9checking how to print strings... print -r
 10checking whether make supports the include directive... yes (GNU style)
 11checking for gcc... gcc
 12checking whether the C compiler works... yes
 13checking for C compiler default output file name... a.out
 14checking for suffix of executables... 
 15checking whether we are cross compiling... no
 16checking for suffix of object files... o
 17checking whether we are using the GNU C compiler... yes
 18checking whether gcc accepts -g... yes
 19checking for gcc option to accept ISO C89... none needed
 20checking whether gcc understands -c and -o together... yes
 21checking dependency style of gcc... gcc3
 22checking for a sed that does not truncate output... /usr/bin/gsed
 23checking for grep that handles long lines and -e... /usr/bin/ggrep
 24checking for egrep... /usr/bin/ggrep -E
 25checking for fgrep... /usr/bin/ggrep -F
 26checking for ld used by gcc... /usr/bin/ld
 27checking if the linker (/usr/bin/ld) is GNU ld... no
 28checking for BSD- or MS-compatible name lister (nm)... /usr/bin/nm -p
 29checking the name lister (/usr/bin/nm -p) interface... BSD nm
 30checking whether ln -s works... yes
 31checking the maximum length of command line arguments... 1572480
 32checking how to convert sparc-sun-solaris2.11 file names to sparc-sun-solaris2.11 format... func_convert_file_noop
 33checking how to convert sparc-sun-solaris2.11 file names to toolchain format... func_convert_file_noop
 34checking for /usr/bin/ld option to reload object files... -r
 35checking for objdump... no
 36checking how to recognize dependent libraries... pass_all
 37checking for dlltool... no
 38checking how to associate runtime and link libraries... print -r --
 39checking for ar... ar
 40checking for archiver [@FILE](/bitcoin-core-secp256k1/contributor/file/) support... no
 41checking for strip... strip
 42checking for ranlib... ranlib
 43checking command to parse /usr/bin/nm -p output from gcc object... ok
 44checking for sysroot... no
 45checking for a working dd... /usr/bin/dd
 46checking how to truncate binary pipes... /usr/bin/dd bs=4096 count=1
 47checking for mt... mt
 48checking if mt is a manifest tool... no
 49checking how to run the C preprocessor... gcc -E
 50checking for ANSI C header files... yes
 51checking for sys/types.h... yes
 52checking for sys/stat.h... yes
 53checking for stdlib.h... yes
 54checking for string.h... yes
 55checking for memory.h... yes
 56checking for strings.h... yes
 57checking for inttypes.h... yes
 58checking for stdint.h... yes
 59checking for unistd.h... yes
 60checking for dlfcn.h... yes
 61checking for objdir... .libs
 62checking if gcc supports -fno-rtti -fno-exceptions... no
 63checking for gcc option to produce PIC... -fPIC -DPIC
 64checking if gcc PIC flag -fPIC -DPIC works... yes
 65checking if gcc static flag -static works... no
 66checking if gcc supports -c -o file.o... yes
 67checking if gcc supports -c -o file.o... (cached) yes
 68checking whether the gcc linker (/usr/bin/ld -64) supports shared libraries... yes
 69checking whether -lc should be explicitly linked in... no
 70checking dynamic linker characteristics... solaris2.11 ld.so
 71checking how to hardcode library paths into programs... immediate
 72checking whether stripping libraries is possible... no
 73checking if libtool supports shared libraries... yes
 74checking whether to build shared libraries... yes
 75checking whether to build static libraries... yes
 76checking whether make supports nested variables... (cached) yes
 77checking for pkg-config... /usr/bin/pkg-config
 78checking pkg-config is at least version 0.9.0... yes
 79checking for ar... /usr/bin/ar
 80checking for ranlib... /usr/bin/ranlib
 81checking for strip... /usr/bin/strip
 82checking for gcc... (cached) gcc
 83checking whether we are using the GNU C compiler... (cached) yes
 84checking whether gcc accepts -g... (cached) yes
 85checking for gcc option to accept ISO C89... (cached) none needed
 86checking whether gcc understands -c and -o together... (cached) yes
 87checking dependency style of gcc... (cached) gcc3
 88checking dependency style of gcc... gcc3
 89checking if gcc supports -Werror=unknown-warning-option... no
 90checking if gcc supports -std=c89 -pedantic -Wno-long-long -Wnested-externs -Wshadow -Wstrict-prototypes -Wundef... yes
 91checking if gcc supports -Wno-overlength-strings... yes
 92checking if gcc supports -Wall... yes
 93checking if gcc supports -Wno-unused-function... yes
 94checking if gcc supports -Wextra... yes
 95checking if gcc supports -Wcast-align... yes
 96checking if gcc supports -Wcast-align=strict... yes
 97checking if gcc supports -Wconditional-uninitialized... no
 98checking if gcc supports -fvisibility=hidden... yes
 99checking for x86_64 assembly availability... no
100checking that generated files are newer than configure... done
101configure: creating ./config.status
102config.status: creating Makefile
103config.status: creating libsecp256k1.pc
104config.status: creating src/libsecp256k1-config.h
105config.status: executing depfiles commands
106config.status: executing libtool commands
107
108Build Options:
109  with external callbacks = no
110  with benchmarks         = yes
111  with tests              = yes
112  with coverage           = no
113  with examples           = no
114  module ecdh             = no
115  module recovery         = no
116  module extrakeys        = no
117  module schnorrsig       = no
118
119  asm                     = no
120  ecmult window size      = 15
121  ecmult gen prec. bits   = 4
122
123  valgrind                = no
124  CC                      = gcc
125  CPPFLAGS                =  
126  SECP_CFLAGS             = -O2  -std=c89 -pedantic -Wno-long-long -Wnested-externs -Wshadow -Wstrict-prototypes -Wundef -Wno-overlength-strings -Wall -Wno-unused-function -Wextra -Wcast-align -Wcast-align=strict -fvisibility=hidden 
127  CFLAGS                  = -g -O2
128  LDFLAGS                 = 
129
130$ nice gmake
131gmake  all-am
132gmake[1]: Entering directory '/export/home/mihailp/secp256k1-master'
133  CC       src/bench.o
134In file included from src/bench.c:11:
135src/util.h:193:43: error: operator '&&' has no right operand
136  193 |      (defined(_BIG_ENDIAN) && _BIG_ENDIAN == 1)
137      |                                           ^~
138src/util.h:198:3: error: #error Please make sure that either SECP256K1_LITTLE_ENDIAN or SECP256K1_BIG_ENDIAN is set, see src/util.h.
139  198 | # error Please make sure that either SECP256K1_LITTLE_ENDIAN or SECP256K1_BIG_ENDIAN is set, see src/util.h.
140      |   ^~~~~
141gmake[1]: *** [Makefile:1130: src/bench.o] Error 1
142gmake[1]: Leaving directory '/export/home/mihailp/secp256k1-master'
143gmake: *** [Makefile:886: all] Error 2

sipa commented at 7:43 pm on February 25, 2022: contributor

Can you try compiling with -DSECP256K1_BIG_ENDIAN, since apparently the autodetection based on macros doesn’t work?
real-or-random commented at 11:14 pm on February 25, 2022: contributor
@tankf33der Can you post the output of this command?
0echo | gcc -dM -E -
This will help us to improve the autodetection, see https://stackoverflow.com/a/2224357.

(Judging from the error message, I assume _BIG_ENDIAN is defined but as the empty string instead of 1…)

tankf33der commented at 5:33 am on February 26, 2022: none

echo | gcc -dM -E -

  0$ echo | gcc -dM -E -
  1#define __DBL_MIN_EXP__ (-1021)
  2#define __UINT_LEAST16_MAX__ 0xffff
  3#define __ATOMIC_ACQUIRE 2
  4#define __FLT128_MAX_10_EXP__ 4932
  5#define __FLT_MIN__ 1.17549435082228750796873653722224568e-38F
  6#define __GCC_IEC_559_COMPLEX 2
  7#define __UINT_LEAST8_TYPE__ unsigned char
  8#define __INTMAX_C(c) c ## L
  9#define __CHAR_BIT__ 8
 10#define __UINT8_MAX__ 0xff
 11#define __WINT_MAX__ 0x7fffffff
 12#define __FLT32_MIN_EXP__ (-125)
 13#define __ORDER_LITTLE_ENDIAN__ 1234
 14#define __SIZE_MAX__ 0xffffffffffffffffUL
 15#define __WCHAR_MAX__ 0x7fffffff
 16#define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_1 1
 17#define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_2 1
 18#define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4 1
 19#define __DBL_DENORM_MIN__ ((double)4.94065645841246544176568792868221372e-324L)
 20#define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8 1
 21#define __GCC_ATOMIC_CHAR_LOCK_FREE 2
 22#define __GCC_IEC_559 2
 23#define __FLT32X_DECIMAL_DIG__ 17
 24#define __FLT_EVAL_METHOD__ 0
 25#define __FLT64_DECIMAL_DIG__ 17
 26#define __GCC_ATOMIC_CHAR32_T_LOCK_FREE 2
 27#define __UINT_FAST64_MAX__ 0xffffffffffffffffUL
 28#define __SIG_ATOMIC_TYPE__ int
 29#define __DBL_MIN_10_EXP__ (-307)
 30#define __FINITE_MATH_ONLY__ 0
 31#define __FLT32X_MAX_EXP__ 1024
 32#define __GNUC_PATCHLEVEL__ 0
 33#define __FLT32_HAS_DENORM__ 1
 34#define __UINT_FAST8_MAX__ 0xff
 35#define __FLT32_MAX_10_EXP__ 38
 36#define __INT8_C(c) c
 37#define __INT_LEAST8_WIDTH__ 8
 38#define sparc 1
 39#define __UINT_LEAST64_MAX__ 0xffffffffffffffffUL
 40#define __SHRT_MAX__ 0x7fff
 41#define __LDBL_MAX__ 1.18973149535723176508575932662800702e+4932L
 42#define __FLT64X_MAX_10_EXP__ 4932
 43#define __FLT64X_HAS_QUIET_NAN__ 1
 44#define __UINT_LEAST8_MAX__ 0xff
 45#define __GCC_ATOMIC_BOOL_LOCK_FREE 2
 46#define __FLT128_DENORM_MIN__ 6.47517511943802511092443895822764655e-4966F128
 47#define __UINTMAX_TYPE__ long unsigned int
 48#define __FLT_EVAL_METHOD_TS_18661_3__ 0
 49#define __unix 1
 50#define __UINT32_MAX__ 0xffffffffU
 51#define __LDBL_MAX_EXP__ 16384
 52#define __FLT128_MIN_EXP__ (-16381)
 53#define __WINT_MIN__ (-__WINT_MAX__ - 1)
 54#define __sun 1
 55#define __FLT128_MIN_10_EXP__ (-4931)
 56#define __INT_LEAST16_WIDTH__ 16
 57#define __SCHAR_MAX__ 0x7f
 58#define __FLT128_MANT_DIG__ 113
 59#define __WCHAR_MIN__ (-__WCHAR_MAX__ - 1)
 60#define __INT64_C(c) c ## L
 61#define __GCC_ATOMIC_POINTER_LOCK_FREE 2
 62#define __SIZEOF_INT__ 4
 63#define __FLT32X_MANT_DIG__ 53
 64#define __USER_LABEL_PREFIX__ 
 65#define __FLT64X_EPSILON__ 1.92592994438723585305597794258492732e-34F64x
 66#define __STDC_HOSTED__ 1
 67#define __DBL_DIG__ 15
 68#define __FLT32_DIG__ 6
 69#define __FLT_EPSILON__ 1.19209289550781250000000000000000000e-7F
 70#define __SHRT_WIDTH__ 16
 71#define __LDBL_MIN__ 3.36210314311209350626267781732175260e-4932L
 72#define __STDC_UTF_16__ 1
 73#define __FLT64X_DENORM_MIN__ 6.47517511943802511092443895822764655e-4966F64x
 74#define __FLT32X_HAS_INFINITY__ 1
 75#define __INT32_MAX__ 0x7fffffff
 76#define __unix__ 1
 77#define __INT_WIDTH__ 32
 78#define __SIZEOF_LONG__ 8
 79#define __UINT16_C(c) c
 80#define __DECIMAL_DIG__ 36
 81#define __FLT64_EPSILON__ 2.22044604925031308084726333618164062e-16F64
 82#define __FLT64X_MIN_10_EXP__ (-4931)
 83#define __LDBL_HAS_QUIET_NAN__ 1
 84#define __FLT64_MANT_DIG__ 53
 85#define __FLT64X_MANT_DIG__ 113
 86#define __GNUC__ 10
 87#define __FLT_HAS_DENORM__ 1
 88#define __SIZEOF_LONG_DOUBLE__ 16
 89#define __BIGGEST_ALIGNMENT__ 16
 90#define __FLT64_MAX_10_EXP__ 308
 91#define __DBL_MAX__ ((double)1.79769313486231570814527423731704357e+308L)
 92#define __INT_FAST32_MAX__ 0x7fffffff
 93#define __DBL_HAS_INFINITY__ 1
 94#define __HAVE_SPECULATION_SAFE_VALUE 1
 95#define __SVR4 1
 96#define __INTPTR_WIDTH__ 64
 97#define __FLT64X_HAS_INFINITY__ 1
 98#define __UINT_LEAST32_MAX__ 0xffffffffU
 99#define __FLT32X_HAS_DENORM__ 1
100#define __INT_FAST16_TYPE__ int
101#define __LDBL_HAS_DENORM__ 1
102#define __FLT128_HAS_INFINITY__ 1
103#define __DBL_MAX_EXP__ 1024
104#define __WCHAR_WIDTH__ 32
105#define __FLT32_MAX__ 3.40282346638528859811704183484516925e+38F32
106#define __GCC_ATOMIC_LONG_LOCK_FREE 2
107#define __PTRDIFF_MAX__ 0x7fffffffffffffffL
108#define __FLT32_HAS_QUIET_NAN__ 1
109#define __LONG_LONG_MAX__ 0x7fffffffffffffffLL
110#define __SIZEOF_SIZE_T__ 8
111#define __sparc__ 1
112#define __FLT64X_MIN_EXP__ (-16381)
113#define __SIZEOF_WINT_T__ 4
114#define __LONG_LONG_WIDTH__ 64
115#define __FLT32_MAX_EXP__ 128
116#define __GXX_ABI_VERSION 1014
117#define __FLT_MIN_EXP__ (-125)
118#define __INT16_MAX__ 0x7fff
119#define __INT_FAST64_TYPE__ long int
120#define __FLT128_NORM_MAX__ 1.18973149535723176508575932662800702e+4932F128
121#define __FLT64_DENORM_MIN__ 4.94065645841246544176568792868221372e-324F64
122#define __DBL_MIN__ ((double)2.22507385850720138309023271733240406e-308L)
123#define __FLT64X_NORM_MAX__ 1.18973149535723176508575932662800702e+4932F64x
124#define __SIZEOF_POINTER__ 8
125#define __LP64__ 1
126#define __DBL_HAS_QUIET_NAN__ 1
127#define __FLT32X_EPSILON__ 2.22044604925031308084726333618164062e-16F32x
128#define __FLT64_MIN_EXP__ (-1021)
129#define __FLT64_MIN_10_EXP__ (-307)
130#define __FLT64X_DECIMAL_DIG__ 36
131#define __REGISTER_PREFIX__ 
132#define __UINT16_MAX__ 0xffff
133#define __DBL_HAS_DENORM__ 1
134#define __FLT32_MIN__ 1.17549435082228750796873653722224568e-38F32
135#define __UINT8_TYPE__ unsigned char
136#define __FLT_DIG__ 6
137#define __NO_INLINE__ 1
138#define __sparcv9 1
139#define __DEC_EVAL_METHOD__ 2
140#define __FLT_MANT_DIG__ 24
141#define __LDBL_DECIMAL_DIG__ 36
142#define __VERSION__ "10.2.0"
143#define __UINT64_C(c) c ## UL
144#define __INT_LEAST32_MAX__ 0x7fffffff
145#define __GCC_ATOMIC_INT_LOCK_FREE 2
146#define __FLT128_MAX_EXP__ 16384
147#define __FLT32_MANT_DIG__ 24
148#define __FLOAT_WORD_ORDER__ __ORDER_BIG_ENDIAN__
149#define __FLT128_HAS_DENORM__ 1
150#define __sparc 1
151#define __FLT128_DIG__ 33
152#define __SCHAR_WIDTH__ 8
153#define __INT32_C(c) c
154#define __ORDER_PDP_ENDIAN__ 3412
155#define sun 1
156#define __INT_FAST32_TYPE__ int
157#define __UINT_LEAST16_TYPE__ short unsigned int
158#define unix 1
159#define __SIZE_TYPE__ long unsigned int
160#define __UINT64_MAX__ 0xffffffffffffffffUL
161#define __FLT64X_DIG__ 33
162#define __INT8_TYPE__ char
163#define __ELF__ 1
164#define __FLT_RADIX__ 2
165#define __INT_LEAST16_TYPE__ short int
166#define __LDBL_EPSILON__ 1.92592994438723585305597794258492732e-34L
167#define __UINTMAX_C(c) c ## UL
168#define __FLT32X_MIN__ 2.22507385850720138309023271733240406e-308F32x
169#define __SIG_ATOMIC_MAX__ 0x7fffffff
170#define __GCC_ATOMIC_WCHAR_T_LOCK_FREE 2
171#define __SIZEOF_PTRDIFF_T__ 8
172#define __LDBL_DIG__ 33
173#define __FLT32X_MIN_EXP__ (-1021)
174#define __INT_FAST16_MAX__ 0x7fffffff
175#define __FLT64_DIG__ 15
176#define __UINT_FAST32_MAX__ 0xffffffffU
177#define __UINT_LEAST64_TYPE__ long unsigned int
178#define __FLT_HAS_QUIET_NAN__ 1
179#define __FLT_MAX_10_EXP__ 38
180#define __LONG_MAX__ 0x7fffffffffffffffL
181#define __FLT64X_HAS_DENORM__ 1
182#define __FLT_HAS_INFINITY__ 1
183#define __UINT_FAST16_TYPE__ unsigned int
184#define __INT_FAST32_WIDTH__ 32
185#define __CHAR16_TYPE__ short unsigned int
186#define __PRAGMA_REDEFINE_EXTNAME 1
187#define __SIZE_WIDTH__ 64
188#define __INT_LEAST16_MAX__ 0x7fff
189#define __INT64_MAX__ 0x7fffffffffffffffL
190#define __FLT32_DENORM_MIN__ 1.40129846432481707092372958328991613e-45F32
191#define __SIG_ATOMIC_WIDTH__ 32
192#define __INT_LEAST64_TYPE__ long int
193#define __INT16_TYPE__ short int
194#define __INT_LEAST8_TYPE__ char
195#define __STDC_VERSION__ 201710L
196#define __INT_FAST8_MAX__ 0x7f
197#define __FLT128_MAX__ 1.18973149535723176508575932662800702e+4932F128
198#define __INTPTR_MAX__ 0x7fffffffffffffffL
199#define __FLT64_HAS_QUIET_NAN__ 1
200#define __FLT32_MIN_10_EXP__ (-37)
201#define __FLT32X_DIG__ 15
202#define __PTRDIFF_WIDTH__ 64
203#define __LDBL_MANT_DIG__ 113
204#define __svr4__ 1
205#define __FLT64_HAS_INFINITY__ 1
206#define __FLT64X_MAX__ 1.18973149535723176508575932662800702e+4932F64x
207#define __SIG_ATOMIC_MIN__ (-__SIG_ATOMIC_MAX__ - 1)
208#define __INTPTR_TYPE__ long int
209#define __UINT16_TYPE__ short unsigned int
210#define __WCHAR_TYPE__ int
211#define __SIZEOF_FLOAT__ 4
212#define __UINTPTR_MAX__ 0xffffffffffffffffUL
213#define __INT_FAST64_WIDTH__ 64
214#define __FLT32_DECIMAL_DIG__ 9
215#define __INT_FAST64_MAX__ 0x7fffffffffffffffL
216#define __GCC_ATOMIC_TEST_AND_SET_TRUEVAL 255
217#define __FLT_NORM_MAX__ 3.40282346638528859811704183484516925e+38F
218#define __FLT32_HAS_INFINITY__ 1
219#define __FLT64X_MAX_EXP__ 16384
220#define __UINT_FAST64_TYPE__ long unsigned int
221#define __INT_MAX__ 0x7fffffff
222#define __INT64_TYPE__ long int
223#define __FLT_MAX_EXP__ 128
224#define __ORDER_BIG_ENDIAN__ 4321
225#define __DBL_MANT_DIG__ 53
226#define __INT_LEAST64_MAX__ 0x7fffffffffffffffL
227#define __GCC_ATOMIC_CHAR16_T_LOCK_FREE 2
228#define __WINT_TYPE__ int
229#define __UINT_LEAST32_TYPE__ unsigned int
230#define __SIZEOF_SHORT__ 2
231#define __FLT32_NORM_MAX__ 3.40282346638528859811704183484516925e+38F32
232#define __LDBL_MIN_EXP__ (-16381)
233#define __FLT64_MAX__ 1.79769313486231570814527423731704357e+308F64
234#define __WINT_WIDTH__ 32
235#define __INT_LEAST8_MAX__ 0x7f
236#define __INT_LEAST64_WIDTH__ 64
237#define __FLT32X_MAX_10_EXP__ 308
238#define __SIZEOF_INT128__ 16
239#define __LDBL_MAX_10_EXP__ 4932
240#define __ATOMIC_RELAXED 0
241#define __DBL_EPSILON__ ((double)2.22044604925031308084726333618164062e-16L)
242#define __FLT128_MIN__ 3.36210314311209350626267781732175260e-4932F128
243#define _LP64 1
244#define __UINT8_C(c) c
245#define __FLT64_MAX_EXP__ 1024
246#define __INT_LEAST32_TYPE__ int
247#define __sun__ 1
248#define __SIZEOF_WCHAR_T__ 4
249#define __UINT64_TYPE__ long unsigned int
250#define __FLT64_NORM_MAX__ 1.79769313486231570814527423731704357e+308F64
251#define __FLT128_HAS_QUIET_NAN__ 1
252#define __INTMAX_MAX__ 0x7fffffffffffffffL
253#define __INT_FAST8_TYPE__ char
254#define __FLT64X_MIN__ 3.36210314311209350626267781732175260e-4932F64x
255#define __LDBL_HAS_INFINITY__ 1
256#define __GNUC_STDC_INLINE__ 1
257#define __FLT64_HAS_DENORM__ 1
258#define __FLT32_EPSILON__ 1.19209289550781250000000000000000000e-7F32
259#define __DBL_DECIMAL_DIG__ 17
260#define __STDC_UTF_32__ 1
261#define __INT_FAST8_WIDTH__ 8
262#define __FLT32X_MAX__ 1.79769313486231570814527423731704357e+308F32x
263#define __DBL_NORM_MAX__ ((double)1.79769313486231570814527423731704357e+308L)
264#define __BYTE_ORDER__ __ORDER_BIG_ENDIAN__
265#define __INTMAX_WIDTH__ 64
266#define __UINT32_C(c) c ## U
267#define __FLT_DENORM_MIN__ 1.40129846432481707092372958328991613e-45F
268#define __INT8_MAX__ 0x7f
269#define __LONG_WIDTH__ 64
270#define __UINT_FAST32_TYPE__ unsigned int
271#define __FLT32X_NORM_MAX__ 1.79769313486231570814527423731704357e+308F32x
272#define __CHAR32_TYPE__ unsigned int
273#define __FLT_MAX__ 3.40282346638528859811704183484516925e+38F
274#define __INT32_TYPE__ int
275#define __SIZEOF_DOUBLE__ 8
276#define __FLT_MIN_10_EXP__ (-37)
277#define __FLT64_MIN__ 2.22507385850720138309023271733240406e-308F64
278#define __INT_LEAST32_WIDTH__ 32
279#define __INTMAX_TYPE__ long int
280#define __FLT32X_HAS_QUIET_NAN__ 1
281#define __ATOMIC_CONSUME 1
282#define __GNUC_MINOR__ 2
283#define __INT_FAST16_WIDTH__ 32
284#define __UINTMAX_MAX__ 0xffffffffffffffffUL
285#define __FLT32X_DENORM_MIN__ 4.94065645841246544176568792868221372e-324F32x
286#define __DBL_MAX_10_EXP__ 308
287#define __LDBL_DENORM_MIN__ 6.47517511943802511092443895822764655e-4966L
288#define __INT16_C(c) c
289#define __PTRDIFF_TYPE__ long int
290#define __ATOMIC_SEQ_CST 5
291#define __arch64__ 1
292#define __UINT32_TYPE__ unsigned int
293#define __FLT32X_MIN_10_EXP__ (-307)
294#define __UINTPTR_TYPE__ long unsigned int
295#define __LDBL_MIN_10_EXP__ (-4931)
296#define __FLT128_EPSILON__ 1.92592994438723585305597794258492732e-34F128
297#define __SIZEOF_LONG_LONG__ 8
298#define __FLT128_DECIMAL_DIG__ 36
299#define __GCC_ATOMIC_LLONG_LOCK_FREE 2
300#define __FLT_DECIMAL_DIG__ 9
301#define __UINT_FAST16_MAX__ 0xffffffffU
302#define __LDBL_NORM_MAX__ 1.18973149535723176508575932662800702e+4932L
303#define __GCC_ATOMIC_SHORT_LOCK_FREE 2
304#define __UINT_FAST8_TYPE__ unsigned char
305#define __ATOMIC_ACQ_REL 4
306#define __ATOMIC_RELEASE 3

tankf33der commented at 5:44 am on February 26, 2022: none

Can you try compiling with -DSECP256K1_BIG_ENDIAN, since apparently the autodetection based on macros doesn’t work?

Helped to compile, tests passed.

 0$ gmake check
 1gmake  check-am
 2gmake[1]: Entering directory '/export/home/mihailp/secp256k1-master'
 3gmake  check-TESTS
 4gmake[2]: Entering directory '/export/home/mihailp/secp256k1-master'
 5gmake[3]: Entering directory '/export/home/mihailp/secp256k1-master'
 6PASS: tests
 7PASS: exhaustive_tests
 8============================================================================
 9Testsuite summary for libsecp256k1 0.1.0-pre
10============================================================================
11# TOTAL: 2
12# PASS:  2
13# SKIP:  0
14# XFAIL: 0
15# FAIL:  0
16# XPASS: 0
17# ERROR: 0
18============================================================================
19gmake[3]: Leaving directory '/export/home/mihailp/secp256k1-master'
20gmake[2]: Leaving directory '/export/home/mihailp/secp256k1-master'
21gmake[1]: Leaving directory '/export/home/mihailp/secp256k1-master'

real-or-random commented at 10:24 am on February 26, 2022: contributor
Thanks for the quick reply. The output does not include _BIG_ENDIAN, so this is probably defined somewhere else.

So the following two are not equivalent if FOO is defined but empty:
0#if defined(FOO) 1#if FOO = 1 2... 3#endif 4#endif
0#if defined(FOO) && FOO == 1 1... 2#endif
We took the code partly from https://github.com/rofl0r/endianness.h/blob/9853923246b065a3b52d2c43835f3819a62c7199/endianness.h#L52L73 and changed it slightly. The original code checks the most commonly used macro __BYTE_ORDER__ separately and then skips all the other checks, so that would have worked here. But still, maybe the entire thing is somewhat fragile. [1] agrees with this observation and favors run-time detection, which then is again turned into compile-time detection by a clever compiler.

The only place where we have endianness-dependant code is when reading input data for SHA256: https://github.com/bitcoin-core/secp256k1/blob/master/src/hash_impl.h#L54

So we could:
1. Stick to the precompiler macros and improve them. This will at least fix this issue here.
2. Switch to runtime detection, e.g., as with a inlineable and compile-time-evalulatable function as suggested in [1]
3. Make the code endianness-independent, preferably in a way that has in the end no overhead for LE.
I think we should try 3 and benchmark if it really has an effect on performance. If not, we should favor that approach. It’s by far the least brittle. Otherwise 2 could be a solid option, too.

For 1, I’m a little bit concerned by the observation in [1] that the macro approach really led to misdetection cases. We have the self-tests but they’re not always run (static context). Unfortunately there seems to be no portable C89 expression that we could use in assumptions.h (Even if compile-time constantness is not a requirement, there’s (*(uint16_t *)"\0\xff" < 0x0100) in C89 but it relies on the assumption that string literals are aligned. The other expressions I found all use C99 compound literals, e.g., see https://stackoverflow.com/a/2100549)

[1] https://fastcompression.blogspot.com/2014/11/portability-woes-endianess-and_24.html
real-or-random referenced this in commit 1f43294e53 on Mar 25, 2022
real-or-random cross-referenced this on Mar 25, 2022 from issue hash: Make code agnostic of endianness by real-or-random
real-or-random referenced this in commit 8d89b9e6e5 on Mar 25, 2022
real-or-random closed this on Mar 28, 2022
real-or-random referenced this in commit 8746600eec on Mar 28, 2022
dderjoel referenced this in commit b9138ab266 on May 23, 2023
matteonardelli referenced this in commit 5ac4bc757f on Jun 16, 2023

Contributors
tankf33der sipa real-or-random