How to override global before_script with empty before_script in job

Question

Is it possible to override the global before_script with an empty before_script within a job?

score 82 · Accepted Answer · answered May 18 '17 at 11:40

82

Yes, the local before_script overrides the global before_script. To keep the .yml sintax valid, use a command that does nothing.

before_script:
- global before script

job:
  before_script:
  - ''

answered May 18 '17 at 11:40

rpadovani

14

How does one do the opposite? Not override the parent `before_script`? – David Callanan Jul 31 '19 at 11:26
@DavidCallanan you can either not declare `before_script` in your job to avoid overriding it or reference the global `before_script` to extend it in your job. https://docs.gitlab.com/ee/ci/yaml/#reference-tags – Antony Fuentes Sep 15 '21 at 17:59
1

To create a reference to the global `before_script`, enter this line into your local `before_script` as if it were a regular command: `- !reference [before_script]`. This way you can not override it but extend it with additional commands. – gekkedev Apr 28 '22 at 09:17
@jakub-kania solution is much "clearer" in my opinion – Izydorr Jan 16 '23 at 11:21

score 74 · Answer 2 · answered May 19 '17 at 05:54

74

You should use an empty array notation.

before_script:
- global before script

job:
  before_script: []
  script:
  - test

answered May 19 '17 at 05:54

Jakub Kania

This appears to simply run the global script instead. – nCrazed Oct 12 '18 at 10:11
@nCrazed You mean the global before_script? It really shouldn't. Please check if your pasted it correctly. – Jakub Kania Oct 23 '18 at 20:37
Yes, I mean the global before_script. – nCrazed Oct 24 '18 at 21:17
@nCrazed Please check the jobs in this project: https://gitlab.com/blckct/before_script , that project contains a demo and it appears to be working just like it should. – Jakub Kania Oct 25 '18 at 07:59
I've been testing it locally via `gitlab-runner` (11.3.1), could that explain the discrepancy in the behavior? :| – nCrazed Oct 26 '18 at 08:23
@nCrazed If you can confirm problem locally you should report it to Gitlab. – Jakub Kania Jan 24 '19 at 11:50
1

Works for me, and syntacticaly I like it better than assigning an empty string as in the accepted answer (which also works) – Auspex Nov 24 '22 at 12:42

2 Answers2