robots-txt | page may be crawled from robots.txt , robots meta tags | Sitemap library

CSS modules in Gatsby v3 onwards needs to be imported as ES modules, so your:

After some debugging, the issue seems to be related to the Contentful source plugin (gatsby-source-contentful) according to some GitHub threads and to the capability to create internal IDs for tag node (tags___NODE).

Aside of waiting the resolution you can try updating the plugin to the latest version.

When I moved to Vercel hosting the problem was solved. Before that I used Netlify.

but i dont know what is the node version on github actions i can not reproduce the error in my local, because of the version are not the same

You could use setup-node action to make the version exactly same with your local:

Try running gatsby clean first, and then try it again…

There's not much debug in the question however, to me, it seems that you are using some dependencies outside React's scope, which may potentially block React's hydration process, which may explain the problem described. For example, using Leaflet instead of React's Leaflet or (depending on its implementation) Bootstrap instead of React's Boostrap.

You should be able to change all React-unfriendly modules to React's ones without too much effort and that should fix your problems.

Keep in mind that if your project "work in develop and breaks in build" doesn't mean that your project work or stops working, it just means that is working under certain and specific circumstances. Basically, and summarizing (to avoid extending the answer), gatsby develop uses the browser as an interpreter, where there are, among other things, global objects like window or document. However, gatsby build occurs in the Node server, where at the build time, there are not global objects because there are not even created yet, that the main reason why you may face a different behavior between both scenarios but doesn't mean that the project stopped working magically.

You can read further details in the Overview of Gatsby Build Process.

Another option, linked with blocking React's hydration, is that some component may be blocking that process because of its own behavior. Be careful when using global objects like window or document (or when importing modules that uses it), they use should be always be wrapped inside the following condition, as you can see from the docs:

When referencing window in a React component.

What did you try so far? As @ksav pointed out, in this GitHub thread there are a several ways to fix a similar issue:

• Removing node_modules, .cache and install again
• Removing node_modules, .cache fix Gatsby to v2.23.3/upgrade up to ^2.26.1 where the bug is fixed and install again

It seems related to a loading staticQuery bug that can't be reproduced in a fresh install. The final trial is to remove your package-lock/yarn-lock.json and generate it again.

After playing around with a few ideas, here is the solution I came other with. Since this question is asked frequently and often opened as a bug against ReadTheDocs on GitHub (which it's not, it just appears to be poorly supported and/or documented), I'll share my workaround here for others to find.

As mentioned above and in the docs, while ReadTheDocs allows you to override the auto-generated robots.txt and publish your own, you can't with sitemap.xml. Unclear why. Regardless, you can simply publish a different sitemap.xml, I named mine sitemap-index.xml, then, tell your robots.txt to point to your custom sitemap.

For my custom sitemap-index.xml, I only put the pages I care about rather then ever generated version (since stable and latest are really what I want search engines to be crawling, not versioned pages):

Smoking Gun was a plugin I was using

"embla-carousel" "embla-carousel-react"

Rolled them back and issues went away, raising an issue on github for the team