Hi, I'm compiling GiNaC 1.0.9 with gcc3.1, and everything seems to go ok, no error messages from 'make'. But 'make check' - which also compiles fine - fails on all three tests. CLN tests ran fine. Any ideas? Thanks, Will make[2]: Entering directory `/home/will/tmp/GiNaC-1.0.9/check' GiNaC will now take an exam with specific input (like a pupils' exam): examining several historic failures just out of paranoia.............. passed examining consistency of numeric types....../run_exams: line 3: 27225 Aborted ./exams 2>exams.out cmp: EOF on exams.out FAIL: run_exams GiNaC will now run through some rather costly random consistency checks: checking consistency of numeric types.. passed checking consistency of symbolic functions.... passed checking symbolic matrix manipulations..... passed checking linear solve....../run_checks: line 3: 27249 Aborted ./checks 2>checks.out cmp: EOF on checks.out FAIL: run_checks GiNaC will now run through some basic timings: timing commutative expansion and substitution.... passed size: 25 50 100 200 time/s: 0.01 0.04 0.19 0.88 timing Laurent series expansion of Gamma function./run_times: line 3: 27277 Aborted ./times 2>times.out cmp: EOF on times.out FAIL: run_times =================== 3 of 3 tests failed =================== make[2]: *** [check-TESTS] Error 1 make[2]: Leaving directory `/home/will/tmp/GiNaC-1.0.9/check' make[1]: *** [check-am] Error 2 make[1]: Leaving directory `/home/will/tmp/GiNaC-1.0.9/check' make: *** [check-recursive] Error 1 __________________________________________________ Do You Yahoo!? Everything you'll ever need on one web page from News and Sport to Email and Music Charts http://uk.my.yahoo.com
On Tue, 16 Jul 2002, William Woods wrote:
I'm compiling GiNaC 1.0.9 with gcc3.1, and everything seems to go ok, no error messages from 'make'. But 'make check' - which also compiles fine - fails on all three tests. CLN tests ran fine. Any ideas?
Thanks,
Will
make[2]: Entering directory `/home/will/tmp/GiNaC-1.0.9/check' GiNaC will now take an exam with specific input (like a pupils' exam): examining several historic failures just out of paranoia.............. passed examining consistency of numeric types....../run_exams: line 3: 27225 Aborted ./exams 2>exams.out cmp: EOF on exams.out FAIL: run_exams GiNaC will now run through some rather costly random consistency checks: checking consistency of numeric types.. passed checking consistency of symbolic functions.... passed checking symbolic matrix manipulations..... passed checking linear solve....../run_checks: line 3: 27249 Aborted ./checks 2>checks.out cmp: EOF on checks.out FAIL: run_checks GiNaC will now run through some basic timings: timing commutative expansion and substitution.... passed size: 25 50 100 200 time/s: 0.01 0.04 0.19 0.88 timing Laurent series expansion of Gamma function./run_times: line 3: 27277 Aborted ./times 2>times.out cmp: EOF on times.out FAIL: run_times =================== 3 of 3 tests failed =================== make[2]: *** [check-TESTS] Error 1 make[2]: Leaving directory `/home/will/tmp/GiNaC-1.0.9/check' make[1]: *** [check-am] Error 2 make[1]: Leaving directory `/home/will/tmp/GiNaC-1.0.9/check' make: *** [check-recursive] Error 1
Judging from the pattern of failures it seems that it crashes whenever an exception is thrown. Probably exceptions go uncatched, for whatever reason. In any case, CLN 1.1.5 and GiNaC 1.0.x have been tested with GCC 3.1 on a variety of platforms! Please tell us exactly what platform you have, which compiler-flags you passed, how you configured and if there was anything suspicious. I'll try to reproduce it then. Regards -richy. -- Richard B. Kreckel <Richard.Kreckel@Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/>
I have rerun 'make check' several times, and one time it only failed on 2 tests (it passed the 'run_checks'). Perhaps this is a hardware fault of some kind?? Platform: Mandrake 8.2 Linux gcc 3.1 CLN 1.1.5 GiNaC 1.0.9 2000+ Athlon 1.5GB DDRAM I did: ./configure --prefix=/usr/local2 make make check No unusual errors or warnings were reported during compilation. Will --- "Richard B. Kreckel" <kreckel@thep.physik.uni-mainz.de> wrote: > On Tue, 16 Jul 2002, William Woods wrote:
I'm compiling GiNaC 1.0.9 with gcc3.1, and everything seems to go ok, no error messages from 'make'. But 'make check' - which also compiles fine - fails on all three tests. CLN tests ran fine. Any ideas?
Thanks,
Will
make[2]: Entering directory `/home/will/tmp/GiNaC-1.0.9/check' GiNaC will now take an exam with specific input (like a pupils' exam): examining several historic failures just out of paranoia.............. passed examining consistency of numeric types....../run_exams: line 3: 27225 Aborted
./exams 2>exams.out cmp: EOF on exams.out FAIL: run_exams GiNaC will now run through some rather costly
consistency checks: checking consistency of numeric types.. passed checking consistency of symbolic functions....
random passed
checking symbolic matrix manipulations..... passed checking linear solve....../run_checks: line 3: 27249 Aborted ./checks 2>checks.out cmp: EOF on checks.out FAIL: run_checks GiNaC will now run through some basic timings: timing commutative expansion and substitution.... passed size: 25 50 100 200 time/s: 0.01 0.04 0.19 0.88 timing Laurent series expansion of Gamma function./run_times: line 3: 27277 Aborted ./times 2>times.out cmp: EOF on times.out FAIL: run_times =================== 3 of 3 tests failed =================== make[2]: *** [check-TESTS] Error 1 make[2]: Leaving directory `/home/will/tmp/GiNaC-1.0.9/check' make[1]: *** [check-am] Error 2 make[1]: Leaving directory `/home/will/tmp/GiNaC-1.0.9/check' make: *** [check-recursive] Error 1
Judging from the pattern of failures it seems that it crashes whenever an exception is thrown. Probably exceptions go uncatched, for whatever reason.
In any case, CLN 1.1.5 and GiNaC 1.0.x have been tested with GCC 3.1 on a variety of platforms! Please tell us exactly what platform you have, which compiler-flags you passed, how you configured and if there was anything suspicious. I'll try to reproduce it then.
Regards -richy. -- Richard B. Kreckel <Richard.Kreckel@Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/>
__________________________________________________ Do You Yahoo!? Everything you'll ever need on one web page from News and Sport to Email and Music Charts http://uk.my.yahoo.com
On Tue, 16 Jul 2002, William Woods wrote:
I have rerun 'make check' several times, and one time it only failed on 2 tests (it passed the 'run_checks'). Perhaps this is a hardware fault of some kind??
Platform:
Mandrake 8.2 Linux
Uhh, anybody else out there with such a Mandrake system?
gcc 3.1 CLN 1.1.5 GiNaC 1.0.9 2000+ Athlon 1.5GB DDRAM
I did: ./configure --prefix=/usr/local2 make make check
No unusual errors or warnings were reported during compilation.
What were the compiler flags ($CXXFLAGS)? Regards -richy. -- Richard B. Kreckel <Richard.Kreckel@Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/>
The problem seems to be solved by specifying: CXXFLAGS="-DDO_GINAC_ASSERT -ggdb -Wall -O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2 (a variation on what is suggested in the tutorial. Infact the '-pedantic' flag causes ./configure not to find cln for some reason so I left it out) Thanks for your help. Will - "Richard B. Kreckel" <kreckel@thep.physik.uni-mainz.de> wrote: > On Tue, 16 Jul 2002, William Woods wrote:
I have rerun 'make check' several times, and one time it only failed on 2 tests (it passed the 'run_checks'). Perhaps this is a hardware fault of some kind??
Platform:
Mandrake 8.2 Linux
Uhh, anybody else out there with such a Mandrake system?
gcc 3.1 CLN 1.1.5 GiNaC 1.0.9 2000+ Athlon 1.5GB DDRAM
I did: ./configure --prefix=/usr/local2 make make check
No unusual errors or warnings were reported during compilation.
What were the compiler flags ($CXXFLAGS)?
Regards -richy. -- Richard B. Kreckel <Richard.Kreckel@Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/>
__________________________________________________ Do You Yahoo!? Everything you'll ever need on one web page from News and Sport to Email and Music Charts http://uk.my.yahoo.com
On Tue, 16 Jul 2002, William Woods wrote:
The problem seems to be solved by specifying:
CXXFLAGS="-DDO_GINAC_ASSERT -ggdb -Wall -O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2
Could you please try changing some of these flags to reasonable defaults such that we find out which of the original setings (presumably "-g -O2") causes the probem? Or, leave the CXXFLAGS and configure specifying --disable-static because without any of --disable-{shared,static} you were most probably linking against the shared library and EH there is more tricky. So, my guesses are: 1) shared library exception handling is broken on your system, 2) -O2 miscomples some code. In that order. The other flags are most probably irrelevant. Where did you get the compiler from? Did you bootstrap it or did it come with the system? Regards -richy. -- Richard B. Kreckel <Richard.Kreckel@Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/>
Ok, I tried some different variations: CXXFLAGS="-finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2 passed all tests, but changing to --disable-static failed all tests. Also, CXXFLAGS="-O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2 failed only on one single test: timing Lewis-Wester test M1 (26x26 sparse, det). failed 0.095s but CXXFLAGS="-DDO_GINAC_ASSERT -O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2 passed all tests. (The gcc3.1 compiler was not standard with this system, I bootstrapped it) Any ideas? Will --- "Richard B. Kreckel" <kreckel@thep.physik.uni-mainz.de> wrote: > On Tue, 16 Jul 2002, William Woods wrote:
The problem seems to be solved by specifying:
CXXFLAGS="-DDO_GINAC_ASSERT -ggdb -Wall -O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2
Could you please try changing some of these flags to reasonable defaults such that we find out which of the original setings (presumably "-g -O2") causes the probem? Or, leave the CXXFLAGS and configure specifying --disable-static because without any of --disable-{shared,static} you were most probably linking against the shared library and EH there is more tricky.
So, my guesses are: 1) shared library exception handling is broken on your system, 2) -O2 miscomples some code. In that order. The other flags are most probably irrelevant.
Where did you get the compiler from? Did you bootstrap it or did it come with the system?
Regards -richy. -- Richard B. Kreckel <Richard.Kreckel@Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/>
__________________________________________________ Do You Yahoo!? Everything you'll ever need on one web page from News and Sport to Email and Music Charts http://uk.my.yahoo.com
On Wed, 17 Jul 2002, William Woods wrote:
Ok, I tried some different variations:
CXXFLAGS="-finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2
passed all tests, but changing to --disable-static failed all tests.
This supports the hypothesis that exceptions are not caught with the shared library. Could you please verify that with the *shared* (.so) library the following program crashes while it runs with the *static* (.a) library? (BTW, it works both ways here---using a self-bootstrapped GCC 3.1 on Debian potato and woody.) ----<tryandcatch.cc>-------------------------------------------------- #include <iostream> #include <ginac/ginac.h> using namespace std; using namespace GiNaC; int main(void) { ex foo; try { foo = 1/foo; } catch (const exception& e) { clog << "caught " << e.what() << endl; } return 0; } ----------------------------------------------------------------------
Also,
CXXFLAGS="-O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2
failed only on one single test:
timing Lewis-Wester test M1 (26x26 sparse, det). failed 0.095s
but
CXXFLAGS="-DDO_GINAC_ASSERT -O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2
passed all tests.
Now that is *very* strange. Did it continue running after M1 failed?
(The gcc3.1 compiler was not standard with this system, I bootstrapped it)
Okay, please send me the output of 'g++ -v' and 'ld -v'
Any ideas?
No, not really (yet). I really wish somebody else could verify this on another current Mandrake system. Regards -richy. -- Richard B. Kreckel <Richard.Kreckel@Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/>
Problem solved - it seems I somehow managed to use gcc3.1 on cln but 2.96 on GiNaC. Everything works fine now. Thanks for all your help. Will --- "Richard B. Kreckel" <kreckel@thep.physik.uni-mainz.de> wrote: > On Wed, 17 Jul 2002, William Woods wrote:
Ok, I tried some different variations:
CXXFLAGS="-finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2
passed all tests, but changing to --disable-static failed all tests.
This supports the hypothesis that exceptions are not caught with the shared library. Could you please verify that with the *shared* (.so) library the following program crashes while it runs with the *static* (.a) library? (BTW, it works both ways here---using a self-bootstrapped GCC 3.1 on Debian potato and woody.)
----<tryandcatch.cc>--------------------------------------------------
#include <iostream> #include <ginac/ginac.h> using namespace std; using namespace GiNaC;
int main(void) { ex foo; try { foo = 1/foo; } catch (const exception& e) { clog << "caught " << e.what() << endl; } return 0; }
----------------------------------------------------------------------
Also,
CXXFLAGS="-O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2
failed only on one single test:
timing Lewis-Wester test M1 (26x26 sparse, det). failed 0.095s
but
CXXFLAGS="-DDO_GINAC_ASSERT -O -finline-limit=1000" ./configure --disable-shared --prefix=/usr/local2
passed all tests.
Now that is *very* strange. Did it continue running after M1 failed?
(The gcc3.1 compiler was not standard with this system, I bootstrapped it)
Okay, please send me the output of 'g++ -v' and 'ld -v'
Any ideas?
No, not really (yet). I really wish somebody else could verify this on another current Mandrake system.
Regards -richy. -- Richard B. Kreckel <Richard.Kreckel@Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/>
__________________________________________________ Do You Yahoo!? Everything you'll ever need on one web page from News and Sport to Email and Music Charts http://uk.my.yahoo.com
participants (2)
-
Richard B. Kreckel
-
William Woods