Le 31/07/2017 à 10:57, Guillaume MM a écrit :
I spotted various areas of improvement:
...
* The new buffer param and the boolean argument in the session file are superfluous and make the code much more complicated than needed.
... I now understand better that this is by design, but it (still) does not seem very natural to me. There are three levels, because there is 1. a switch to enable shell-escape (after which we are still given a further dialog at compilation before actually running with shell-escape). 2. a button on this dialog to further "always allow". The red icon on the bottom right is given when 1. is active even though one is still safe as long as 2. has not been pressed. With e.g. Sweave the first switch is automatic and innocuous (there are Sweave insets, so ask for running sweave). Then with my patch the icon informs when one has authorized to never ask anymore, so one is in an actual unsafe state. With Enrico's patch, somebody using minted will always check the switch 1. and from this point on he always sees the red icon, so the icon provides zero information. The contextual menu reverts the choice of converters, which is pointless because it just makes the document uncompilable. There is no option to see just revoke 2. or to see the status of 2. As a conclusion, as it is, the status bar icon in this patch is not an implementation of visibility nor revocability. For it to be so, the icon should be shown when 2. is active, and the context menu should switch 2. instead of 1. If furthermore a validation mechanism was provided for minted requiring shell-escape, then one would just need 2 levels as with needauth.