Re: [gecode-users] Bug with Tuple, Minimizescript (Restart) and several threads

[Manuel Baclet]

Hi Guido, thanks for the fix. I have found two other problems (using osX 10.8). When possible, i tried with Gecode 3.7.3 and the trunk version. - EPK_SPEED does not work with restart search and several threads. I provide an example:

test.cp
Description: Binary data

This gives me seg faults. gdb says: Program received signal EXC_BAD_ACCESS, Could not access memory. Reason: KERN_INVALID_ADDRESS at address: 0x0000000000000008 [Switching to process 8791 thread 0x1203] 0x000000010004cbcd in Gecode::Int::Extensional::Incremental<Gecode::Int::BoolView>::remove_support (this=0x103008990, home=@0x102800000, l=0x101308280, i=1, n=1) at incremental.hpp:254 254      while ((*a)->t != l) { - sequence is not working properly with restart search and several threads. An example:

test2.cp
Description: Binary data

This gives different random error messages including Segmentation fault: 11, Abort trap: 6 and  <<test2(8717,0x10d484000) malloc: *** error for object 0x7fb3c3c081d0: pointer being freed was not allocated      *** set a breakpoint in malloc_error_break to debug>> gdb reports one of the following error (depending on the execution): Program received signal EXC_BAD_ACCESS, Could not access memory. Reason: 13 at address: 0x0000000000000000 [Switching to process 8828 thread 0x1203] 0x00000001001a1f78 in Gecode::IntSet::~IntSet () at core.hpp:2286 2286      delete o; or  Program received signal EXC_BAD_ACCESS, Could not access memory. Reason: KERN_INVALID_ADDRESS at address: 0x0000000000000000 [Switching to process 8834 thread 0x1203] Gecode::VarImp<Gecode::Int::IntVarImpConf>::advise (this=<value temporarily unavailable, due to optimizations>, home=@0x1013086d0, me=1, d=@0x1012cee20) at core.hpp:3346 3346      switch (p.advise(home,*a,d)) { or test2(8866,0x1012cf000) malloc: *** error for object 0x1013081f0: pointer being freed was not allocated *** set a breakpoint in malloc_error_break to debug Program received signal SIGABRT, Aborted. [Switching to process 8866 thread 0x1203] 0x00007fff916e2212 in __pthread_kill () The following example shows that this can also happen with bab search:

test3.cp
Description: Binary data

Hope you can fix this, Cheers, Manuel Le 19 juil. 2012 à 08:55, Guido Tack <t...@gecode.org> a écrit : Hi Manuel, the bug is now fixed in the svn trunk.  Thanks again for reporting it! Cheers, Guido On 16/07/2012, at 8:30 AM, Manuel Baclet wrote: Hello, first of all, thank you for this great software tool. I think there is a bug in Gecode when using extensional constraint with tuples,  minimizescript, restart search and several threads. For example, the following program terminates most of the time with a segmentation fault. I tested this on osX and Linux: <test.cp> Everything is fine when using one thread or using bab search and i don't think that the problem occurs when using DFA or regular expressions instead of tuples. On Linux, gdb says: [New Thread 0xb62b6b40 (LWP 7988)] [New Thread 0xb5ab5b40 (LWP 7989)] [New Thread 0xb52b4b40 (LWP 7990)] [New Thread 0xb4ab3b40 (LWP 7991)] Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 0xb62b6b40 (LWP 7988)] 0xb7aacf95 in Gecode::Int::Extensional::Base<Gecode::Int::BoolView, true>::valid (this=0x806f030, t=0x11, dom=0x806e038)   at ./gecode/int/extensional/base.hpp:131 131      if (!dom[i].get(static_cast<unsigned int>(t[i]-ts()->min))) Regards, Manuel _______________________________________________ Gecode users mailing list users@gecode.org https://www.gecode.org/mailman/listinfo/gecode-users

_______________________________________________
Gecode users mailing list
users@gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users