MESSAGE
| DATE | 2016-11-29 |
| FROM | Ruben Safir
|
| SUBJECT | Re: [Hangout-NYLXS] [Learn] Look at this exciting output by my test
|
On 11/29/2016 06:00 PM, Christopher League wrote:
>
> First thing I try to do is to reduce the program to the SMALLEST program
> that reproduces the same error. You can think of it as binary search:
> delete (something like) HALF the program. If error is not reproduced,
> restore that half and delete the OTHER HALF. If error is reproduced,
> then delete half AGAIN. Continue until it's as small as possible. Here's
> the minimal crashing program I came up with:
>
> int main ( int argc, char *argv[] )
> {
> std::vector test {0};
> assert(test.size() == 1); // VECTOR HAS JUST ONE ELEMENT!
> for(int i = 0; i<10; i++ ){
> test[i] = i; // THEREFORE THIS IS AN ERROR WHEN i>0
> }
> return 0;
> }
>
> It exhibits the same error yours does: "free(): invalid next size
> (fast)". I added the `assert` to test my assumption about what the `{0}`
> initializer means.
>
> I tried going even further by unrolling and 'sampling' the loop:
>
> int main ( int argc, char *argv[] )
> {
> std::vector test {0};
> assert(test.size() == 1); // VECTOR HAS JUST ONE ELEMENT!
> test[0] = 0; // SAFE
> test[1] = 1; // ERR
> test[9] = 9; // ERR
> return 0;
> }
>
> But this -- unfortunately, IMO -- does not crash. Nevertheless, It's
> still erroneous. It's ABSOLUTELY NOT ALLOWED to read or write `test[1]`
> or `test[9]` because `test.size() == 1`.
>
> Anyway, the `vector` type HAS a dereference operation that does
> bounds-checking. You'll be much better off if you USE IT. That would
> look like this:
>
> int main ( int argc, char *argv[] )
> {
> std::vector test {0};
> assert(test.size() == 1); // VECTOR HAS JUST ONE ELEMENT!
> test.at(0) = 0; // SAFE
> test.at(1) = 1; // CRASH!
> test.at(9) = 9; // UNREACHABLE
> return 0;
> }
>
> This time it crashes with a much more helpful error message:
>
> terminate called after throwing an instance of 'std::out_of_range'
> what(): vector::_M_range_check: __n (which is 1) >= this->size() (which is 1)
>
> and if you back-trace in `gdb` it pinpoints exactly where the error is.
> (Unlike with your initial error, when it actually crashes AFTER the
> actual erroneous access.)
>
> Due to the bounds-checking, you haven't overwritten any of the memory
> management record-keeping, so it can give you honest results.
>
> If you don't like bounds-checking, TOUGH... do it anyway. ONCE THE
> PROGRAM IS WORKING, you can optimize it away where you really think you
> need to. (Usually you don't.)
>
> That's my $0.03.
>
> CL
>
maybe this
http://rsos.royalsocietypublishing.org/content/3/3/150636
not what I was thinking of but a good article
--
So many immigrant groups have swept through our town
that Brooklyn, like Atlantis, reaches mythological
proportions in the mind of the world - RI Safir 1998
http://www.mrbrklyn.com
DRM is THEFT - We are the STAKEHOLDERS - RI Safir 2002
http://www.nylxs.com - Leadership Development in Free Software
http://www2.mrbrklyn.com/resources - Unpublished Archive
http://www.coinhangout.com - coins!
http://www.brooklyn-living.com
Being so tracked is for FARM ANIMALS and and extermination camps,
but incompatible with living as a free human being. -RI Safir 2013
_______________________________________________
hangout mailing list
hangout-at-nylxs.com
http://www.nylxs.com/
|
|
