fix: use a direct bash script instead of an intermediary node script to install Cypress in the factory

[cacieprins]

The existing install script can exit with code 0 if the npm command fails, instead of failing. This can cause that docker layer to cache as successful even if it failed. To resolve, the intermediary node.js script is removed, as everything it does can be accomplished in a shell script. Some of the echos are removed, as the command itself should be visible in the docker logs - they are unnecessary.

[MikeMcC399]

@cacieprins

I also saw this and I suggest just to put

set -e

at the top of

https://github.com/cypress-io/cypress-docker-images/blob/master/factory/installScripts/cypress/default.sh

The problem is that if the download fails, then it has already created the /root/.cache directory.

I already had a PR prepared for this, but didn't want to confuse things by submitting it right now!

The problem at the moment is actually the network between CircleCI and Cloudflare. I can build the image locally without any problem.

[MikeMcC399]

• See also Cypress install download error not caught by installation script #1391 which I had written before you posted your PR

[cacieprins]

@MikeMcC399 Thanks!

[MikeMcC399]

@cacieprins

I've submitted #1392 which is a simple fix for the incorrect error reporting.

If this PR succeeds (using 14.5.2) and the 14.5.3 PR fails, then it's likely to be caused by a Cloudflare issue.

[cacieprins]

@MikeMcC399 I'd like to use this one, since it removes a fully redundant intermediary script.

[MikeMcC399]

@cacieprins

I will drop out now. It's up to you and the Cypress.io team to decide what to do.

• PR fix: catch Cypress installation script download error #1392 is a very simple solution to the problem about catching the error properly.