Question

all. I’m working on software to block ads using the system’s hosts file, but on to the code.

I have a custom class HostEntry that contains necessary information such as the destination host, the host to block, the destination host’s ip… etc.

In the HostsManager class, it keeps a vector to keep track of all the hosts added. In order to block a host completely, I must add example.com AND www.example.com, but when I iterate over the vector it will only delete the entry starting with “www.” and leaves the one without. If you try to delete it a second time (with only the entry missing the “www.”) it segfaults, and I don’t know why.

void HostsManager::delHost(std::string blockedhost) {
    strip(blockedhost);
    string tmp; // yes I know it's not great practice to do it like this, but it was for debug reasons
    for (vector<HostEntry>::iterator viter = hosts.begin(); viter != hosts.end(); ++viter) {
        tmp = viter->getHost();
        if (tmp == blockedhost || tmp == ("www." + blockedhost)) {
            viter = hosts.erase(viter);
        }
    }
}

An example call to that specific function:

HostsManager mgr;
mgr.delHost("mysite.com"); // this deletes "www.mysite.com" but not "mysite.com" - whether or not you call delHost() with the "www." prefix
mgr.delHost("mysite.com"); // if you call it a second time, it segfaults O.o

Help with this would be GREATLY appreciated.

EDIT: I assigned the value returned from the call to erase() to viter, same result. I still have no idea why this is happening.

If you need all the code, it’s at http://paste.pocoo.org/show/363051/

Answer 1

It is generally better to use std::remove_if to delete multiple items from a vector; it runs in linear time instead of quadratic and avoids having to worry about iterator invalidation.

Might look something like this:

hosts.erase(std::remove_if(hosts.begin(), 
                           hosts.end(), 
                           [&blockedhost](const HostEntry& entry) {
                               return entry.getHost() == blockedhost || 
                                      entry.getHost() == "www." + blockedhost;
                           }),
            hosts.end());

You can accomplish a similar thing without C++0x lambdas via a struct to compare with:

struct RemoveBlockedHost {
    RemoveBlockedHost(const std::string& s): blockedHost(s) {}
    bool operator () (const HostEntry& entry) {
        return entry.getHost() == blockedHost || entry.getHost() == "www." + blockedHost;
    }
    const std::string& blockedHost;
};

hosts.erase(std::remove_if(hosts.begin(), hosts.end(), RemoveBlockedHost(blockedhost)), hosts.end());