Hi all,

I'm having a really terrible bug that seemed to come from nowhere and is really hard to narrow down. I have a threaded message service that works via local TcpSocket. Every time I run it, either an error saying:
>Unable to open 'recv.c': Unable to read file '/build/glibc-ZN95T4/glibc-2.31/sysdeps/unix/sysv/linux/recv.c' (Error: Unable to resolve non-existing file '/build/glibc-ZN95T4/glibc-2.31/sysdeps/unix/sysv/linux/recv.c').
when socket.receive() is called. Sometimes I get the same thing but for socket.accept() instead (different file). This seems to generate a core.exception.InvalidMemoryOperationError that I can't catch.

Yesterday, this came up after a couple of minutes of the program running, now it happens straight away. I haven't touched this class in a week or so and can't think of why all of a sudden it has come up with this issue. Code dump below

---------------------------------------------


>module message_service;
>
>import std.stdio;
>import std.socket;
>import core.thread;
>
>
>/**
>Threaded socket manager to serve inter-service communication
>
>Calls delegates when the appropriate command is received
>*/
>class MessageService : Thread{
>    /// The socket over which communication is possible
>    private Socket server;
>    /// The connection to the client that sends commands
>    private Socket client;
>    /// Associative array of command strings and relative functions to call
>    private void delegate()[string] commands;
>    /// Global buffer. Holds the command being recieved
>    private char[1024] buffer = 0;
> 
>
>    ///
>    this(ushort port, void delegate()[string] _commands){
>        commands = _commands;
>        server = new TcpSocket();
>        server.setOption(SocketOptionLevel.SOCKET, SocketOption.REUSEADDR, true);
>        server.bind(new InternetAddress(port));
>        writefln("Message service started on port %d", port);
>        super(&mainLoop);
>        start();
>    }
>
>
>    /// Loops over polling the client socket for commands
>    void mainLoop(){
>        while(true){
>            pollMessages();
>        }
>    }
>
>
>    /**
>    Polls the client socket for characters to build up the message buffer
> 
>    When a string command is found, it calls the appropriate function
>    */
>    void pollMessages(){
>        server.listen(1);
>        if(client is null) client = server.accept();
>
>        // Add to the message buffer
>        auto buffLength = client.receive(buffer);
>
>        if(buffLength){
>            auto message = cast(string) buffer[0 .. buffLength];
> 
>            foreach(cmd; commands.byKey){
>                if(cmd == message){
>                    // Reset the message buffer
>                    buffer = 0;
>                    commands[cmd]();
>                    return;
>                }
>            }
>        }
>    }
>}

On Wednesday, 20 January 2021 at 21:31:54 UTC, Tim wrote:
>>Unable to open 'recv.c': Unable to read file '/build/glibc-ZN95T4/glibc-2.31/sysdeps/unix/sysv/linux/recv.c' (Error: Unable to resolve non-existing file '/build/glibc-ZN95T4/glibc-2.31/sysdeps/unix/sysv/linux/recv.c').
> [snip]
> generate a core.exception.InvalidMemoryOperationError that I can't catch.


None of this makes much sense given the code you provided. InvalidMemoryOperationError (the scariest thing in D btw, such a pain to debug) is usually caused by a class destructor somewhere, and none of those should be trying to resolve those files.

Do you have any destructors defined?

On Thursday, 21 January 2021 at 03:21:41 UTC, Adam D. Ruppe wrote:
> On Wednesday, 20 January 2021 at 21:31:54 UTC, Tim wrote:
>>>Unable to open 'recv.c': Unable to read file '/build/glibc-ZN95T4/glibc-2.31/sysdeps/unix/sysv/linux/recv.c' (Error: Unable to resolve non-existing file '/build/glibc-ZN95T4/glibc-2.31/sysdeps/unix/sysv/linux/recv.c').
>> [snip]
>> generate a core.exception.InvalidMemoryOperationError that I can't catch.
>
>
> None of this makes much sense given the code you provided. InvalidMemoryOperationError (the scariest thing in D btw, such a pain to debug) is usually caused by a class destructor somewhere, and none of those should be trying to resolve those files.
>
> Do you have any destructors defined?

No, I don't. It should be all garbage collected right?

Hahahaha, about 50% of my troubles are InvalidMemoryOperationError! It's so frustrating. This whole project has been a nightmare

On Thursday, 21 January 2021 at 03:21:41 UTC, Adam D. Ruppe wrote:
> On Wednesday, 20 January 2021 at 21:31:54 UTC, Tim wrote:
>>>Unable to open 'recv.c': Unable to read file '/build/glibc-ZN95T4/glibc-2.31/sysdeps/unix/sysv/linux/recv.c' (Error: Unable to resolve non-existing file '/build/glibc-ZN95T4/glibc-2.31/sysdeps/unix/sysv/linux/recv.c').
>> [snip]
>> generate a core.exception.InvalidMemoryOperationError that I can't catch.
>
>
> None of this makes much sense given the code you provided. InvalidMemoryOperationError (the scariest thing in D btw, such a pain to debug) is usually caused by a class destructor somewhere, and none of those should be trying to resolve those files.
>
> Do you have any destructors defined?

You seem like a rather switched-on fellow. Would you be able to send me an email at some point at tim.oliver@tutanota.com. I have a proposition for you.

On Thursday, 21 January 2021 at 03:24:13 UTC, Tim wrote:
> No, I don't. It should be all garbage collected right?

Yeah, but that's where the problem comes.

Note that by destructor, I mean *any* function in your code called `~this() {}`. If there are any and they call a memory allocation function, that's how you get the InvalidMemoryOperationError.

I'd have to look up if there's any other causes of that...

On Thursday, 21 January 2021 at 03:35:05 UTC, Adam D. Ruppe wrote:
> On Thursday, 21 January 2021 at 03:24:13 UTC, Tim wrote:
>> No, I don't. It should be all garbage collected right?
>
> Yeah, but that's where the problem comes.
>
> Note that by destructor, I mean *any* function in your code called `~this() {}`. If there are any and they call a memory allocation function, that's how you get the InvalidMemoryOperationError.
>
> I'd have to look up if there's any other causes of that...

I am aware of this, there is absolutely no ~this in *my* code

On Thursday, 21 January 2021 at 03:30:50 UTC, Tim wrote:
> On Thursday, 21 January 2021 at 03:21:41 UTC, Adam D. Ruppe wrote:
>> On Wednesday, 20 January 2021 at 21:31:54 UTC, Tim wrote:
>>>>[...]
>>> [snip]
>>> generate a core.exception.InvalidMemoryOperationError that I can't catch.
>>
>>
>> None of this makes much sense given the code you provided. InvalidMemoryOperationError (the scariest thing in D btw, such a pain to debug) is usually caused by a class destructor somewhere, and none of those should be trying to resolve those files.
>>
>> Do you have any destructors defined?
>
> You seem like a rather switched-on fellow. Would you be able to send me an email at some point at tim.oliver@tutanota.com. I have a proposition for you.

On Discord fyi

On Wednesday, 20 January 2021 at 21:31:54 UTC, Tim wrote:
> Hi all,
>
> I'm having a really terrible bug that seemed to come from nowhere and is really hard to narrow down.

You may be hitting this issue:

https://issues.dlang.org/show_bug.cgi?id=7349

FWIW, popped onto the radar in the most recent Foundation meeting.