Thanks! I now understand what you mean by the second call to symbol-value not being needed. The previous behavior only chose this "branch" of the cond if symbol-value was not nil. To keep this behavior but only have one call to symbol-value, why not change to (keep the symbol-value in the cond as opposed to the body of the branch): ((and (symbolp file) (boundp file) (symbol-value file))) to keep the old behavior of the cond statement? YL On Fri, May 19, 2017 at 7:16 AM, Nicolas Goaziou wrote: > Hello, > > Yuri Lensky writes: > > > The second call is indeed needed. > > AFAICT, it isn't. > > (cond ('foo) (t nil)) => 'foo > > > This is the case fixed by the patch: > > > > (setq org-default-notes-file (expand-file-name "~/docs/notes.org")) > > (setq org-capture-templates '(("t" "Todo" entry (file+olp > > org-default-notes-file "Inbox") "* TODO %?\n%i"))) > > OK, I see. I introduced this regression in > 88a3c2483ee47b342e9bb7d2c1645dce11179bf5. > > I applied your patch with a slight change: > > ((and (symbolp file) (boundp file) (symbol-value file)) (symbol-value > file)) > > => > > ((and (symbolp file) (boundp file)) (symbol-value file)) > > Thank you. > > Regards, > > -- > Nicolas Goaziou >