Need help with a segfault problem

Qjay

@jsulm said in Need help with a segfault problem:

void MultiSegmentLogic::deallocate()
{
if (segs) {
delete[] segs;
segs = nullptr;
}
}

hey @jsulm thanks for quick reply but there is still same error

Program received signal SIGSEGV, Segmentation fault. 0x00401983 in MultiSegmentLogic::deallocate (this=0x61fe7c) at deepmem.cpp:234 234 delete[] segs; (gdb)

jsulm

@Qjay Then please put a break point at delete[] segs; and see the status (segs pointing to valid chunk of memory?)

Paul Colby

I'd strongly recommend initialising segs to nullptr in the MultiSegmentLogic constructor. Your current logic certainly allows segs to be accessed uninitialised, given certain values for numSegs.

Try:

MultiSegmentLogic::MultiSegmentLogic(int numSegs = 0) : numSegs(0), segs(nullptr)
{
   ...
}

Qjay

hey , i tried still same error

here is updated code : https://ghostbin.com/paste/q8d5e

changes


MultiSegmentLogic::MultiSegmentLogic(int numSegs = 0) : numSegs(0), segs(nullptr){
    
   if (!setNumSegs(numSegs))
      numSegs = 0;
}

Paul Colby

Good. Now you also have to do the same in BooleanFunc::deallocate() as @jsulm suggested for MultiSegmentLogic::deallocate(), ie

void BooleanFunc::deallocate()
{
    if (truthTable) {
        delete[] truthTable;
        truthTable = nullptr;
    }
}

This is related to the crash point, because the delete in MultiSegmentLogic invokes the BooleanFunc destructors, and those destructors in turn call BooleanFunc::deallocate().

Also, don't forget to set the breakpoint as @jsulm suggested. After that, try valgrind.

Cheers.

Qjay

@jsulm i did breakpoint with this

break MultiSegmentLogic::deallocate

output

Breakpoint 1, MultiSegmentLogic::deallocate (this=0x61fe6c) at deepmem.cpp:234
234         if (segs) {
(gdb) step
235             delete[] segs;
(gdb) step

Program received signal SIGSEGV, Segmentation fault.
0x00401995 in MultiSegmentLogic::deallocate (this=0x61fe6c) at deepmem.cpp:235
235             delete[] segs;
(gdb)

Qjay

@Paul-Colby

updated code here : https://ghostbin.com/paste/cvfq7

changes

void BooleanFunc::deallocate()
{
    if (truthTable) {
        delete[] truthTable;
        truthTable = nullptr;
    }
}

Paul Colby

The (next) problem is with your copy constructor. (valgrind show'd it immediately)

> g++ foo.cpp && valgrind ./a.out
...
==9657== Conditional jump or move depends on uninitialised value(s)
==9657==    at 0x109314: MultiSegmentLogic::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109405: MultiSegmentLogic::operator=(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x1093D8: MultiSegmentLogic::MultiSegmentLogic(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109FB3: main (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657== 
==9657== Conditional jump or move depends on uninitialised value(s)
==9657==    at 0x109325: MultiSegmentLogic::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109405: MultiSegmentLogic::operator=(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x1093D8: MultiSegmentLogic::MultiSegmentLogic(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109FB3: main (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657== 
==9657== Use of uninitialised value of size 8
==9657==    at 0x10933B: MultiSegmentLogic::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109405: MultiSegmentLogic::operator=(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x1093D8: MultiSegmentLogic::MultiSegmentLogic(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109FB3: main (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657== 
==9657== Conditional jump or move depends on uninitialised value(s)
==9657==    at 0x10935A: MultiSegmentLogic::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109405: MultiSegmentLogic::operator=(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x1093D8: MultiSegmentLogic::MultiSegmentLogic(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109FB3: main (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657== 
==9657== Use of uninitialised value of size 8
==9657==    at 0x108F76: BooleanFunc::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x108CFF: BooleanFunc::~BooleanFunc() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109367: MultiSegmentLogic::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109405: MultiSegmentLogic::operator=(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x1093D8: MultiSegmentLogic::MultiSegmentLogic(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109FB3: main (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657== 
==9657== Invalid read of size 8
==9657==    at 0x108F76: BooleanFunc::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x108CFF: BooleanFunc::~BooleanFunc() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109367: MultiSegmentLogic::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109405: MultiSegmentLogic::operator=(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x1093D8: MultiSegmentLogic::MultiSegmentLogic(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109FB3: main (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==  Address 0x7fbb2fe8 is not stack'd, malloc'd or (recently) free'd
==9657== 
==9657== 
==9657== Process terminating with default action of signal 11 (SIGSEGV)
==9657==  Access not within mapped region at address 0x7FBB2FE8
==9657==    at 0x108F76: BooleanFunc::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x108CFF: BooleanFunc::~BooleanFunc() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109367: MultiSegmentLogic::deallocate() (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109405: MultiSegmentLogic::operator=(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x1093D8: MultiSegmentLogic::MultiSegmentLogic(MultiSegmentLogic const&) (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==    by 0x109FB3: main (in /home/paul/src/extern/forum-qt-io/segfault/a.out)
==9657==  If you believe this happened as a result of a stack
==9657==  overflow in your program's main thread (unlikely but
==9657==  possible), you can try to increase the size of the
==9657==  main thread stack using the --main-stacksize= flag.
==9657==  The main thread stack size used in this run was 8388608.
test 1test 2==9657== 
==9657== HEAP SUMMARY:
==9657==     in use at exit: 357 bytes in 8 blocks
==9657==   total heap usage: 11 allocs, 4 frees, 74,101 bytes allocated
==9657== 
==9657== LEAK SUMMARY:
==9657==    definitely lost: 0 bytes in 0 blocks
==9657==    indirectly lost: 0 bytes in 0 blocks
==9657==      possibly lost: 0 bytes in 0 blocks
==9657==    still reachable: 357 bytes in 8 blocks
==9657==                       of which reachable via heuristic:
==9657==                         newarray           : 80 bytes in 1 blocks
==9657==         suppressed: 0 bytes in 0 blocks
==9657== Rerun with --leak-check=full to see details of leaked memory
==9657== 
==9657== For counts of detected and suppressed errors, rerun with: -v
==9657== Use --track-origins=yes to see where uninitialised values come from
==9657== ERROR SUMMARY: 10 errors from 10 contexts (suppressed: 0 from 0)

At a quick glance it looks like your copy constructor is deallocating before initialising segs.

Cheers.

PS I highly recommend you explore valgrind - such a handy tool for things like this.

Qjay

this section right . but the deallocate() will only be called if this != rhs . still not sure how to fix it

MultiSegmentLogic::MultiSegmentLogic(const MultiSegmentLogic &rhs)
{
   *this = rhs;
}
MultiSegmentLogic & MultiSegmentLogic::operator=(const MultiSegmentLogic &rhs)
{
   if (this != &rhs)
   {
      deallocate();
      this->numSegs = rhs.numSegs;
      allocateSegsArray(numSegs);
      for (int i = 0; i < numSegs; i++)
         this->segs[i] = rhs.segs[i];
   }

   return *this;
}

Thanks i will look and learn more about valgrind .

Qjay

alright i got it .

thanks @Paul-Colby , @jsulm .

the problem was indeed deallocate() . i have removed it . program completes without any errors

aha_1980

Hi @jsulm and @Paul-Colby,

if (segs) {
    delete[] segs;
    segs = nullptr;
}

This is an anti-pattern! Just make sure segs is nullptr before you call new and set it to nullptr after delete. No need to check for nullptr before delete:

delete segs;
segs = nullptr;

Quoting the C++ Reference:

"If this is a null-pointer, the function does nothing."

Regards