the major issue is that we use the build-args to generate the name of the ECR repository we're going to create

which means that we now don't know hte name of the repo because the name itself is dependent on data which is now an output.

Happens here: https://github.com/pulumi/pulumi-cloud/blob/master/aws/service.ts#L162

args are used here: https://github.com/pulumi/pulumi-cloud/blob/master/aws/service.ts#L133-L152

so there's no 'easy' fix here where i could just allow for htis and have it pushed into the next release.

we baked htis pretty deep. but maybe there's some other way of producing the ecr repo name that we could use instead?

For example,

let nginxFargateService = new infra.x.ecs.FargateService("nginx", {
    cluster: cluster,
    taskDefinitionArgs: {
        containers: {
            nginx: {
                image: infra.x.ecs.Image.fromDockerBuild({
                    cacheFrom: true,
                    context: "./nginx",
                }),
                portMappings: nginxLoadBalancer,
                memory: 256,
                environment: [{
                    name: "FOO",
                    value: "foo"
                }]
            }
        }
    },
    desiredCount: 1
});

Is

FOO

available when building the container?

it somewhat depends on what you mean by 'building'. which parts of the dockerfile would need this?

and nginx needs this information at build time?

it cannot use the data when it launches by reading something out of hte environment?

For example,

events {
    worker_connections 1024;
}

http {
    server {
        listen 443;
        ssl_certificate /etc/ssl/certs/nginx-selfsigned.crt;
        ssl_certificate_key /etc/ssl/private/nginx-selfsigned.key;
        ssl_protocols TLSv1.2 TLSv1.1 TLSv1;

        location /api {
            proxy_pass {{api_url}};
            proxy_ssl_server_name on;
            proxy_ssl_protocols TLSv1 TLSv1.1 TLSv1.2;
        }

        location / {
            proxy_pass {{frontend_url}};
            proxy_ssl_protocols TLSv1 TLSv1.1 TLSv1.2;
        }
    }
}

I need to fill in

{{api_url}}

and

{{frontend_url}}

with the values generated from Pulumi resources.

This is currently how I am doing it:

ARG API_URL
ARG FRONTEND_URL
RUN sed -i "s,{{api_url}},"$API_URL",g" /etc/nginx/nginx.conf
RUN sed -i "s,{{frontend_url}},"$FRONTEND_URL",g" /etc/nginx/nginx.conf

also, i'm not opposed to having 'build args' be something that could be the result of a previous resource. it just means though that we need some sort of answer about how we can name these repositories.

i.e. if we allowed build-args to be computed inputs, then the client (i.e. you, would likely have to supply a name for the repo that htey wanted to use)

I can't use `Output`s in

environment

.