On Monday, April 24, 2017 at 1:16:44 PM UTC-4, Anton Gorev wrote: > > Yes, -xmemalign=4i fixed all errors I saw before. >
OK, thanks. -xmemalign=4i appears to be valid for SPARC, and not i.86pc. > But it isn't so cloudless. Next bugreport attached. > > Also maybe output below will be useful: > > RSA validation suite running... > > signal SEGV (no mapping at the fault address) in _memcpy at > 0xffffffff7e200678 > 0xffffffff7e200678: _memcpy+0x0044: ldx [%o1 + %o0], %o3 > Current function is CryptoPP::CopyWords > 28 memcpy(r, a, n*WORD_SIZE);^M > (dbx) print r > dbx: internal error: > /home/abuild/agorev/cryptopp/cryptopp-master.2/misc.h:51: bad token '\' > > dbx: warning: Errors encountered while skimming "integer.cpp" for macro > definitions. > See `help macro skimming-errors' for more explanation. > This appears to be a bug in DBX. Also see https://stackoverflow.com/questions/43624030/is-backslash-newline-valid-on-if-and-elif. > > r = 0x1011a3c40 > (dbx) print a > a = (nil) > (dbx) list - > 18 for (size_t i=0; i<n; i++)^M > 19 r[i] = a;^M > 20 }^M > 21 ^M > 22 inline void CopyWords(word *r, const word *a, size_t n)^M > 23 {^M > 24 if (r != a)^M > 25 #if CRYPTOPP_MSC_VERSION^M > 26 memcpy_s(r, n*WORD_SIZE, a, n*WORD_SIZE);^M > 27 #else^M > (dbx) up > Current function is CryptoPP::Integer::Integer > 2972 CopyWords(reg, t.reg, reg.size());^M > (dbx) print t > t = { > reg = { > m_alloc = { > /* try using "print -r" to see any inherited members */ > } > m_mark = 1025U > m_size = 0 > m_ptr = (nil) > } > sign = POSITIVE > } > (dbx) list - > 2962 ^M > 2963 Integer::Integer()^M > 2964 : reg(2), sign(POSITIVE)^M > 2965 {^M > 2966 reg[0] = reg[1] = 0;^M > 2967 }^M > 2968 ^M > 2969 Integer::Integer(const Integer& t)^M > 2970 : reg(RoundupSize(t.WordCount())), sign(t.sign)^M > 2971 {^M > (dbx) up > Current function is CryptoPP::RSAFunction::PreimageBound > 47 Integer PreimageBound() const {return m_n;}^M > (dbx) print m_n > dbx: internal error: > /home/abuild/agorev/cryptopp/cryptopp-master.2/misc.h:51: bad token '\' > > dbx: warning: Errors encountered while skimming "rsa.cpp" for macro > definitions. > See `help macro skimming-errors' for more explanation. > > m_n = { > reg = { > m_alloc = { > /* try using "print -r" to see any inherited members */ > } > m_mark = 1025U > m_size = 0 > m_ptr = (nil) > } > sign = POSITIVE > } > (dbx) > > Nothing is jumping out at me for the RSA gear. For completeness, did you perform a 'make distclean' to ensure everything got rebuilt using the same compiler flags? Maybe a debug build will help isolate the issue. The asserts may alert to something fishy. To perform a debug build: make distclean CXX=/opt/.../bin/CC CXXFLAGS="-DDEBUG -g2 -xO0" gmake -j 4 > P.S. Changes in the CMakeLists.txt: > > $ diff -u ../../cryptopp-master/CMakeLists.txt ../CMakeLists.txt > --- ../../cryptopp-master/CMakeLists.txt Tue Apr 18 07:57:38 2017 > +++ ../CMakeLists.txt Mon Apr 24 17:21:07 2017 > @@ -205,6 +205,8 @@ > > # Solaris specific > if ((NOT CRYPTOPP_CROSS_COMPILE) AND "${UNAME_SYSTEM}" STREQUAL "SunOS") > + set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -xmemalign=4i") > + > if (NOT DISABLE_NATIVE_ARCH) > # SunCC needs -native > if ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "SunPro") > $ > > I think that change casts a wider net than needed. -xmemalign=4i should be limited to SPARC64, and maybe SPARC. It should not apply to the Intel machines. Applying it to Intel machines will cause a warning. > Do I need to prepare pull request for this change? And should I revert > commit with workaround that we merged before (as it is not necessary > because -xmemalign=4i fixes original bugs)? > Yeah, we may as well back out the former change, and go with the -xmemalign=4i. A new PR will be fine. Jeff -- -- You received this message because you are subscribed to the "Crypto++ Users" Google Group. To unsubscribe, send an email to cryptopp-users-unsubscr...@googlegroups.com. More information about Crypto++ and this group is available at http://www.cryptopp.com. --- You received this message because you are subscribed to the Google Groups "Crypto++ Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to cryptopp-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
