Re: [Boost-users] Unordered map copy segfaults! (STL map does not)

Subject: Re: [Boost-users] Unordered map copy segfaults! (STL map does not)
From: MeMooMeM (mbelgin_at_[hidden])
Date: 2009-08-21 18:18:45

MeMooMeM wrote:
> 
> 
> Mathias Gaunard-2 wrote:
>> 
>> You should write a proper testcase and provide your platform, so that 
>> people have a chance to reproduce your results.
>> 
> 
> That's right. I prepared a toy code as you suggested, but it works just
> fine (?!). When I use the same structure inside my code, it segfaults. It
> is most probably due to some messed up memory section. Still it is strange
> that I get no segfaults (also no detected errors by valgrind) using STL's
> map...
> 
> Everyone, please ignore this post until I find a proper way to demonstrate
> the problem. 
> 
> Thanks!
> 

OK, mystery is solved. Here's what happened, in case someone else has this
problem:

I was *not* using the standard for loop, as in:

for (Map_t::iterator it = somemap.begin(); it != somemap.end(); it++)

because I was erasing elements from the map, which is known to mess things
up if a for loop is being used. Instead, I was using a while loop and
incrementing the iterator in a more controlled way:

somemap.erase(it++);

which erases the element but still returns correclty incremented iterator.
However forgot to check for the case: (it++ == somemap.end()) in the while
loop, hence the segmentation faults. 

Thanks again!

 

-- 
View this message in context: http://www.nabble.com/Unordered-map-copy-segfaults%21-%28STL-map-does-not%29-tp25034074p25088176.html
Sent from the Boost - Users mailing list archive at Nabble.com.