Fix potential Null pointer dereference in test-fdleak.c

[icy17]

No description provided.

[azat]

fd leaked here

[icy17]

OK, I'll fix it

[azat]

I wish that the main would have a normal error path, i.e.:

if (!start_client)
    goto cleanup;
...
cleanup:
    if (listener)
        evconnlistener_free(listener);
    if (base)
        event_base_free(base);

[icy17]

OK, thanks for the example

[icy17]

I previously misunderstood; while modifying the code, I noticed that I hadn't altered the main function, so perhaps adding a cleanup path is unnecessary? Also, regarding the 'fd leaked' issue, there seems to be no file descriptor here. I believe there are no newly allocated variables within this function, thus no need for cleanup. If I've overlooked something, please highlight it for me.

[azat]

I noticed that I hadn't altered the main function, so perhaps adding a cleanup path is unnecessary

You didn't, but you modified start_client, and main calls it (start_loop -> start_client), so if it can fail it should propagate error to main, since the whole point of this PR to make avoid any leaks on exit AFAIU

Also, regarding the 'fd leaked' issue, there seems to be no file descriptor here

listener_accept_cb accepts socket, if the bufferevent fails you need to close the socket

[icy17]

You didn't, but you modified start_client, and main calls it (start_loop -> start_client), so if it can fail it should propagate error to main, since the whole point of this PR to make avoid any leaks on exit AFAIU

Perhaps I should handle the cleanup in start_loop not in main? Also, I’m not sure whether to change the exit to return in start_client. By doing so, resources could be properly released if start_client fails.

[azat]

Perhaps I should handle the cleanup in start_loop not in main?

This is OK

Also, I’m not sure whether to change the exit to return in start_client. By doing so, resources could be properly released if start_client fails.

Yes, you should not call exit from start_client