X-Debbugs-Cc: joaotavora@gmail.com 1. Have a project with just one file, test.go (attached). 2. Visit it, enable go-ts-mode, call 'M-x eglot'. 3. Move point right after "math.As", press C-M-i. 4. Code will get completed to "math.Asin" This is a problem because "math.As" has more completions (one can look at them by pressing TAB after "math.A") - such as "Acos", "Acosh" and "Abs" - in other words, "fuzzy" matches. Expected behavior: 1. When input is "math.As" - keep the string as-is. 2. When input is "math.Asi" - complete to "math.Asin". This problem seems older than 65ea742ed5ec (the change that introduced eglot--dumb-tryc itself, bug#68699), but it doesn't reproduce with Eglot from Emacs 29. Branches emacs-30 and master are affected. There is also another issue there: when there are no completions at all, this style still has completion-try-completion return a non-nil value (the last line of the implementation). Both of these issues is something I came across when working on closer 'try-completion' integration for company-mode (https://github.com/company-mode/company-mode/pull/1488) and testing out the new code with Eglot. Not sure what's the best fix, but the patch below seems to address both problems in my limited testing. WDYT? diff --git a/lisp/progmodes/eglot.el b/lisp/progmodes/eglot.el index 353877f60c2..e8823a9d2f0 100644 --- a/lisp/progmodes/eglot.el +++ b/lisp/progmodes/eglot.el @@ -3142,8 +3142,14 @@ eglot--dumb-allc (defun eglot--dumb-tryc (pat table pred point) (let ((probe (funcall table pat pred nil))) (cond ((eq probe t) t) - (probe (cons probe (length probe))) - (t (cons pat point))))) + ((and probe + (cl-every + (lambda (s) (string-prefix-p probe s completion-ignore-case)) + (funcall table pat pred t))) + (cons probe (length probe))) + (t + (and (funcall table pat pred t) + (cons pat point)))))) (add-to-list 'completion-category-defaults '(eglot-capf (styles eglot--dumb-flex))) (add-to-list 'completion-styles-alist '(eglot--dumb-flex eglot--dumb-tryc eglot--dumb-allc))