Rendering theme with anything other than "hugo server" does not work?

[mreferre]

I was able to test locally my blog to use the hugo-clarity theme. When serving the blog using hugo server --disableFastRender my new look and feel works just fine:

As I started to migrate it to my staging S3 bucket I have noticed that the theme breaks entirely. I was able to reproduce the problem locally by simply building the site (with the hugo command), move into the public folder generated by the build and serving it using NGINX (with docker run --rm -p 8080:80 -v $(pwd):/usr/share/nginx/html nginx).

This is what I get on either the NGINX local instance (or on the S3 bucket as a try to copy the public folder over):

Using the current theme I can just take the public folder and either serve it with NGINX or via S3 web hosting and it just works. Is there something I am missing to do for the hugo-clarity theme?

At the exact time of this writing this staging S3 bucket has the broken version of my blog per the above: http://wildrydes-mreferre.s3-website-eu-west-1.amazonaws.com/

[mreferre]

I sense it may be due to some CORS issues but I am not entirely sure. Still looking into this.

[mreferre]

Ok it looks like that the baseurl is what's breaking it. If I change the baseurl in my config file to http://wildrydes-mreferre.s3-website-eu-west-1.amazonaws.com/ and I deploy it to my staging bucket above it works fine. Probably related to this issue. I am trying to wrap my head around how to be able to test this on multiple sites (that are not configured to respond with the baseurl such as a staging bucket)...

[mreferre]

Ok... this was it. And I have just figured that the workaround was in the comment above the baseurl entry in the config.toml file.

# if you set it to "/" share icons won't work properly on production
baseurl = "/"  # Include trailing slash

I don't think I have "share icons" in my blog (and the advantages of the workaround are worth it anyway).

UPDATE: it looks like that setting the baseurl to "/" breaks search as well. If true, I will be considering keeping it set to the url of my blog (https://it20.info) and accepting breaking the staging deployments.

[mreferre]

More details: when the blog is deployed with the baseurl set to "/" the search does not work at all and this is what the dev console shows:

[mreferre]

I eventually modified my deployment pipeline (which is configured with a BUCKET variable) to sed the config.toml file. I have configured the config file to set https://it20.info/ as the baseurl variable and if the BUCKET is != than the production bucket I sed the file with the bucket URL:

if [ $BUCKET != "it20.info" ]; then sed -i "s@https://it20.info/@http://${BUCKET}.s3-website-us-east-1.amazonaws.com/@" config/_default/config.toml; fi

In general this has been harder to configure than I thought because of the baseurl sensitivity. I still have one of the redirects (https://www.it20.info) that doesn't work because the www.it20.info bucket I am using to redirect to the main bucket doesn't support https and so it hangs at the "can't connect" before even trying to redirect (my speculation).

I love this theme but this baseurl seems to be hard to get wholly right :)