February 2014 - CLN-list

MinGW64 build
by Robert Szalai 05 Nov '19

05 Nov '19

Hi all, I've been trying to build CLN from windows 64, with some success. It compiles fine, but does not work. The problem is that WIN64 is LLP64, and not ILP64 as Linux. CLN uses long to store pointer data, but size_t or ssize_t should be used as it is guaranteed to be the size of the pointer everywhere. The attached patch was my attempt. Could somebody tell me where else should I look? Pointers are also encoding some metadata in the bits that are set to zero by alingnment, what is this data? There is also some hashing of pointers. Thanks, Robert

6 21

Windows 64-Bit
by Christoph Jung 06 Mar '14

06 Mar '14

Dear Mr Kreckel, I tried to compile the CLN for Windows 64-Bit using x64, but during make I get an error: Type char* does not fit into a long!! What could I do to prevent this from happening? For 32-Bit it worked. Thank you very much and best regards, Christoph Jung Dr. Carl Christoph Jung Robert-Bosch-Zentrum für Leistungselektronik Hochschule Reutlingen Oferdinger Str. 50 72768 Reutlingen-Rommelsbach Tel.: 07121-271-7069

2 3

Dear all
by Christoph Jung 21 Feb '14

21 Feb '14

Dear all, I tried to compile the CLN for Windows 64-Bit using x64, but during make I get an error: Type char* does not fit into a long!! What could I do to prevent this from happening? For 32-Bit it worked. Thank you very much and best regards, Christoph Jung

1 0

Re: [CLN-list] Trouble compiling under cygwin-64
by Richard B. Kreckel 06 Feb '14

06 Feb '14

Hi Pascal, On 02/06/2014 06:33 PM, Pascal Dupuis wrote: > I'm trying to revive the old "symbolic" package from Octave; which > itself is built atop Ginac and CLN. > > So the first task is to compile and install CLN, which fails. I > identified two issues: > > 1) there seems to be a typo in config.h.in. The first line seems to > re-use some symbol defined elsewhere. I modified it as > > #ifndef _CONFIG_H > #define _CONFIG_H You mean, _CL_CONFIG_PUBLIC_H is defined elsewhere? I think you are experiencing interferences with another installation of CLN. > 2) having done this, I re-launched autoreconf. During the configure > stage, the symbol "GMP_DEMANDS_UINTD_LONG" is defined, although it > seems the correct version in cygwin-64 is GMP_DEMANDS_UINTD_LONG_LONG. Can you, please, have a look at the test program emitted by m4/gmp.m4 and see why it gets this wrong? > After this configure step, I modified the generated > include/cln/config.h to ensure the later definition is effective. Okay. I think it's best to fix top issues first. > After steps 1 and 2, the compilation runs over a few files, and then fails with > libtool: compile: g++ -DHAVE_CONFIG_H -I. -I../../src -I../autoconf > -I../../include -I../../src -I../include -I../src -g -O2 -MT > base/digit/cl_2D_div.lo -MD -MP -MF base/digit/.deps/cl_2D_div.Tpo -c > ../../src/base/digit/cl_2D_div.cc -DDLL_EXPORT -DPIC -o > base/digit/.libs/cl_2D_div.o > In file included from ../../src/base/digit/cl_2D.h:7:0, > from ../../src/base/digit/cl_2D_div.cc:7: > ../../src/base/digit/cl_D.h: In function 'sint32 cln::sign_of_sintD(sintD)': > ../../src/base/digit/cl_D.h:35:21: error: call of overloaded > 'sign_of(sintD&)' is ambiguous > return sign_of(wert); > ^ > ../../src/base/digit/cl_D.h:35:21: note: candidates are: > In file included from ../../src/base/digit/cl_D.h:7:0, > from ../../src/base/digit/cl_2D.h:7, > from ../../src/base/digit/cl_2D_div.cc:7: > ../../src/base/cl_low.h:12:15: note: sint16 cln::sign_of(sint16) > inline sint16 sign_of (sint16 wert) > ^ > ../../src/base/cl_low.h:27:15: note: sint32 cln::sign_of(sint32) > inline sint32 sign_of (sint32 wert) > ^ > ../../src/base/cl_low.h:44:15: note: sint64 cln::sign_of(sint64) > inline sint64 sign_of (sint64 wert) > ^ > Makefile:5572: recipe for target 'base/digit/cl_2D_div.lo' failed > make[1]: *** [base/digit/cl_2D_div.lo] Error 1 > make[1]: Leaving directory > '/cygdrive/c/Users/PascalDupuis/Dropbox/Development/cln/compile/src' > Makefile:666: recipe for target 'all-recursive' failed > make: *** [all-recursive] Error 1 > > Although with the updated version, sintD is defined as long long, the > compiler seems unable to pick up the right implementation. Any idea > how to solve this ? I'm taking this to the mailing list cln-list(a)ginac.de. Please subscribe to it. I'm afraid, I cannot help you any better since I don't know anything about cygwin-64. But maybe someone on that list can offer some advice. Best wishes -rbk. -- Richard B. Kreckel <http://www.ginac.de/~kreckel/>

1 0