[patch] Fix #silent macro runs

[psprint]

Hi,
mc has a functionality of skipping the subshell when running a macro script – by prepending the macro with a "#silent" line, e.g.:

#silent

m       make
        TMPFILE=`mktemp /tmp/make.XX` || exit 1
        make 2> $TMPFILE
        mcedit $TMPFILE
        #rm $TMPFILE

This will make the script to be run without the subshell use. However, it's fairly broken currently – the screen will be disrupted – \r and \n codes will not work properly and will not make the caret return to the first column. This is probably also broken also for a non #silent background command runs, however I don't have the opportunity to test it as my subshell support segfaults for some reason.

The patch fixes this problem. It restores the pre-slang/pre-curses terminal state and clears the terminal before running the command, then sets the after-slang/curses terminal configuration.

[andrew_b]

    /* Emit the clear screen escape code. */
    fputs("\x1b[0m\x1b[2J",stdout);

We have ESC_STR.
Why is this sequence used?
​https://stackoverflow.com/questions/37774983/clearing-the-screen-by-printing-a-character

[psprint]

I've removed the screen clearing as it looks even better without it. Thanks for info about ESC_STR.

[andrew_b]

Perhaps, it should be made not for editor only.
Something like following:

src/usermenu.c

@@ -559 +559 @@
 
         if (show_prompt)
             shell_execute (cmd, EXECUTE_HIDE);
-        else if (system (cmd) == -1)
-            message (D_ERROR, MSG_ERROR, "%s", _("Error calling program"));
+        else
+        {
+            gboolean ok;
+
+            /* Prepare the terminal by setting its flag to the initial ones. This will cause \r to work as
+             * expected, instead of being ignored. */
+            tty_reset_shell_mode();
+
+            ok = (system (cmd) != -1);
+
+            /* Restore the SLang terminal configuration and redraw the editor. */
+            tty_raw_mode();
+
+            if (!ok)
+                message (D_ERROR, MSG_ERROR, "%s", _("Error calling program"));
+        }
 
         g_free (cmd);
     }

[psprint]

Yes, and the patch almost works – it only needs a clr_scr(); repaint_screen() calls after the raw mode restoration, because otherwise the screen remains as the external command left it. I attach a tested patch.

[psprint]

The mccore-wide version of the patch.

[andrew_b]

Branch: 4171_silent_macro
changeset:8177cdd0f1668df703f92dfd47034ba9027e16e6

[andrew_b]

Merged to master: [c7306aae18ee3cc1a6ffa659e3ce247ec8744c62].