Ihor Radchenko <yanta...@posteo.net> writes: > Björn Bidar <bjorn.bi...@thaodan.de> writes: > >>> Then, why not simply the attached patch? >>> >> Because the patch circumvents the other-frame mechanism when the >> other-frame is dead. > > May it be that `org-link-frame-setup' is configured by the user to use > current frame, but the user manually used gnus-other-frame and closed > the frame? In my mind, your patch then introduced unexpected behavior.
The `org-link-frame-setup' introduced the first unexpected behavior I think since `org-gnus-follow-link' unconditionally uses the previously existing frame, going with that opening Gnus outside of the gnus-other-frame-object in the frame setup function would go against that. If the user created a frame setup function which used the current frame then the current frame would become the Gnus other-frame as there can be one only Gnus frame. So the patch wouldn't lead to unexpected behavior even for such a user, it's more likely the user would be confused that the Gnus frame was changed to be the current one. >> ... Further the org-link-frame-setup doesn't take >> gnus-other-frame into account either. > > Why would it need to take it into account? The org-gnus-follow-link function tries to take into account whenever the user previously used `gnus-other-frame' to open Gnus by selecting the appropriate frame. However the current org-link-frame-setup function doesn't do so and calls the `gnus' function which would not use the previously existing Gnus frame. The correct approach is to call gnus-other-frame to activate Gnus when `gnus-other-frame-object' exist since that won't activate Gnus in the current window but in the previous or a new frame. The side effect is that if there's a gnus-other-frame-object that is always alive when `org-gnus-follow-link' tries to select it.
