rt-thread | open source IoT real-time operating system

The problem appears to be that the compiler-generated IEnumerable type implements IEnumerable and IEnumerator explicitly:

You can't overwrite the $using: variable reference - but you can use it to dereference a variable value in the calling scope, at which point you can then mutate it (assuming a reference-type value was assigned to the original variable):

You are trying to pass nullptr to the std::thread constructor in B::B():

std::thread doesn't let you pass by reference unless you're explicit because it's an easy source of lifetime issues. Use std::ref to be explicit that you're passing i by reference:

The this pointer needs to be passed as the 1st argument:

It could be reference related.

It is.

Do I still need to wrap something in std::ref() for some reason?

Yes. You need to use std::ref.

Around what?

Around column.

wouldn't wrapping it with std::ref make it a reference to a reference?

There can't be a std::reference_wrapper of a reference. The template will be deduced as std::reference_wrapper> which is what you need.

It works! Could I ask, why exactly is the std::ref() needed there?

Because, as per documentation, std::thread makes copies of its arguments (by moving if possible). In case of columns which is an lvalue to a vector, a copy of the vector will be made. The copies are passed into the function as rvalues. Non-const lvalue references (such as the argument is) do not bind to rvalues, so the program is ill-formed.

what is it accomplishing?

std::reference_wrapper is a copyable object that maintains indirection internally and implicitly converts to lvalue of the referred type, which allows it to be passed into reference arguments. Instead of making a copy of the referred vector, the thread will make a copy of the reference wrapper.

what did you mean by the template being deduced here?

std::ref is a function template. You don't need to explicitly specify its template argument because it can be deduced from the argument that is passed into it.

P.S. Be very careful to keep the referred vector alive at least as long as the thread runs.

So it's taken a week of trial and error to arrive at this point and on the whole I'm pretty happy with the results. The script I'm going to share below takes care of ~3 steps in the processing of the files I'm working with:

1. Creates folders
2. Copies files to new folders
3. Verifies files have copied without error

It does this whilst taking <1/3 of the time that doing steps 1) & 2) in Excel (using FileSystemObject to copy files)

This is not an answer with a solution. But I think I'm reading something that I have experienced on several occasions.

I've been using python and Hashicorp vault to try and manage tokens/RBAC on applications. But very often it would break because it had not updated yet, due to the propagation from AAD to back end being asynchronous from what I was told.

I even did checks where I used ADAL to loop over the application to verify if it was good. But even then it would still fail on some occasions. Which hurt the automation I was trying to put in place.

Now you are having some issue that seems similar, but instead while adding the reply url to an existing application. My question for testing is; does the reply URL work when it is supplied upon creation of the application? If so, and testing is 100%, then you are having the same issue.

For me, pre-creation of all necessary properties on applications is what helped me circumvent this annoying issue. As I don't think adding a sleep anywhere is a good way to move forward, and the reply from the API isn't reliable enough to work on. If pre-creation is not an option, I suppose the sleep timer is probably some way forward. For me, that ended up being 2-5m in some cases. And in some lucky cases 7-30s

In current keycloak docker image (6.0.1), the default stack is UDP. According to this, version 7.0.0 will default to TCP and will also introduce a variable to toggle the stack (JGROUPS_TRANSPORT_STACK).

Using the UDP stack in Amazon ECS will "partially" work, meaning the discovery will work, the cluster will form, but the Infinispan cache won't be able to sync between instances, which will produce erratic errors. There is probably a way to make it work "as-is", but I dont see anything blocked between the instances when checking the VPC Flow logs.

A workaround is to switch to TCP by modifying the JGroups stack directly in the image in file /opt/jboss/keycloak/standalone/configuration/standalone-ha.xml: