labeled_graph::remove_vertex does not remove label

[arneboe]

When calling remove_vertex on a labeled_graph the vertex is removed but the label is not.
A dangling reference to the vertex is left behind. Accessing it causes segfaults.

Minimal example:

#include <iostream>
#include <string>

#include <boost/graph/directed_graph.hpp>
#include <boost/graph/labeled_graph.hpp>

using namespace boost;
using namespace std;

int main() {

    using namespace boost::graph_detail;
    typedef directed_graph<> Digraph;
    typedef labeled_graph<Digraph, string> Graph;
    
    Graph g;
    g.add_vertex("foo");
    
    auto v = g.vertex("foo");
    if(v != nullptr)
        cout << "foo exists\n";
    
    g.remove_vertex("foo");
    
    auto v2 = g.vertex("foo");
    if(v2 != nullptr)
        std::cout << " BUG! vertex label still exists.\n";


    return 0;
}

[anadon]

@jzmaddock Hey, this looks like a good one for me to work on.

[jzmaddock]

Go for it!

[anadon]

@arneboe Can I get your full name and email, since what you submitted is what I'm basing the regression test on? I need to credit you.

[arneboe]

I dont care about the credit, but if you have to: Arne Böckmann

[anadon]

It looks like the intended design, given the state of BGL, is to have _map a protected member, not a private member. So you hack was less of a hack than was expected.

[anadon]

This may take longer than expected: https://github.com/boostorg/graph/blob/develop/include/boost/graph/adjacency_matrix.hpp#L992

[anadon]

May be related: https://svn.boost.org/trac10/ticket/7863

[anadon]

This is going to be delayed somewhat. There are general cleanups in testing which I need to do first.

[HWiese1980]

Sure, take your time! Better have it fixed correctly and thoroughly, than to rush it and only come up with some quick'n'dirty patch that fails somewhere else.

[anadon]

Just an update, had to move twice, once across states, and started a new job. Getting back to this more now.

[HWiese1980]

Alright, no pressure. Just don't forget it ;-)

[fjch1997]

And... it was forgotten
This bug really tripped me up when I tried to use labeled_graph

[HWiese1980]

As it seems... @anadon how's it going?

[jeremy-murphy]

Not sure if they are still active on Boost.Graph, but if someone submits a pull request to fix this issue, I can help them get it polished and merge it.

[EduardoGoulart1]

Hey there, any update on it? @anadon by any chance do you still have your WIP PR that people can pick up from?

[DeuceBox]

This may be related:

There was a patch at https://svn.boost.org/trac/boost/ticket/9493 (I did not create it) which is no longer reachable, but I've been using it in production for years

I've preserved it here:
DeuceBox@de651cb

It appears to fix the bug in the minimal example given in the OP.

[StahlTim]

@DeuceBox @jeremy-murphy
Is there currently an active pull request introducing the fix referenced by @DeuceBox?

Facing the same issue and would appreciate the fix being shipped with the stock boost library.

[DeuceBox]

@StahlTim @jeremy-murphy since there wasn't any real discussion beyond the branch/fix I linked a while back, I've put in a PR.

[jeremy-murphy]

I'll prioritise merging in the fix asap.

[jeremy-murphy]

Could someone please check out the fix I just merged and test it? Thanks to @DeuceBox for making the PR.

[EduardoGoulart1]

Hi @jeremy-murphy, has this fix already been merged somewhere? I can't find it in the latest release

[DeuceBox]

@EduardoGoulart1 It's merged here #354 in the "develop" branch. The boostorg/graph "develop" branch is what the boostorg/boost "master" branch references. However, the latest boostorg/boost release branch/tag (boost-1.85.0) references the corresponding release tag in boostorg/graph (https://github.com/boostorg/graph/releases/tag/boost-1.85.0), but it appears based on https://github.com/boostorg/graph/tags that the boostorg/graph repo's release tags haven't pointed to anything new since 2022.

@jeremy-murphy could we get the next boostorg/graph release tag to reference a more recent version of develop? I don't know what's involved in that process.

[jeremy-murphy]

Thanks for pointing it out. That's about how long I've been maintainer for and I'm still learning what needs to be done! I assumed all that release tag business would be automatic from the super project.