@yyoncho could you have a look here? Without this fix the yasnippet integration in Spacemacs is broken for many layers.

Can someone review this PR? In case you need some details, let me clarify the issue and the new logic proposed by this PR.

The original code intends to add company-capf to company-backends if it doesn't exist in the list. To check the existence of company-capf, equal is used to check each element of company-backends:

(setq-local company-backends (cl-adjoin 'company-capf company-backends :test #'equal)))

However, the use of equal is not appropriate here, because company-backends can contain a sublist as described in https://company-mode.github.io/manual/Backends.html#Grouped-Backends.

For example, if company-capf is included in a grouped backend in company-backends like this:

(setq company-backends '((company-capf :with company-yasnippet)))

the original code fails to detect company-capf inside the sublist and the resulting company-backends has two company-capf like this:

(company-capf (company-capf :with company-yasnippet))

To avoid this issue, the check needs to be done based on type of an element in company-backends: if the element is an atom, check if it is company-capf, if the element is a list, check if campany-capf is a member of the list.

The PR uses the correct logic and I've confirmed that the issue is resolved with it.

The work-around for this is to set lsp-completion-provider to :none and then your company-backends configuration will be kept intact.

Thank you for your suggestion, but I still would like to the PR to be merged because:

1. the original code's assumption that company-backends includes only atoms is incorrect
2. configuring lsp-completion-provider to :none causes an unwanted side effect: company-mode is not enabled via lsp-mode. When lsp-completion-provider is :none, (company-mode 1) at line 918 is not executed: