Output to stderr should be captured on run and displayed as pop-up (configurable)

[zaytsev]

Hi!

Forwarded from Debian: ​http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=581175

From: martin f krafft <madduck@debian.org>
Subject: redirect stderr of helpers
Date: Tue, 11 May 2010 13:17:39 +0200

Many GUI programmes, such as G* and K* tools, spam stderr with crap
that noone cares about. If mc spawns these helpers, the output
messes up the ncurses UI and requires me to punch ctrl-l to
restore.

I think it would make sense if mc categorically intercepted stderr
of processes it spawns. One could consider accumulating such output
in a buffer that can be viewed separately, or to display the output
with e.g. less, and restore the UI on exit. Ideally it would be
configurable, of course. I don't really want to have to hit any keys
to return to the mc UI.

I think there is a room for discussion, whether this has to be implemented or not and if yes, how. If the consensus is that this will never be implemented I will close the downstream bug.

Z.

[gotar]

Something could be done, especially with long-running and 'chatty' programs that keep polluting terminal. However I wonder if it's not easier just to add 2>/dev/null in place where one defines such helpers, thus having the ability to supress some of them while saving others.

And for the ones that are kept I'd like them to clutter only a part of the screen (let's say from the half to mini status) without shifting vertically panels, so I won't have to press ctrl-l to get reliable navigation back.

[zaytsev]

Related to #2354.

[zaytsev]

Related to #2331.

[angel_il]

maybe try this?

diff --git a/lib/vfs/mc-vfs/fish.c b/lib/vfs/mc-vfs/fish.c
index 14b3cf4..ba8a4c0 100644
--- a/lib/vfs/mc-vfs/fish.c
+++ b/lib/vfs/mc-vfs/fish.c
@@ -552,6 +552,7 @@ fish_dir_load (struct vfs_class *me, struct vfs_s_inode *dir, char *remote_path)
     int reply_code;
     gchar *shell_commands;

+    refresh_cmd ();
     /*
      * Simple FISH debug interface :]
      */
diff --git a/src/dialog.c b/src/dialog.c
index a19d39a..ab19cc7 100644
--- a/src/dialog.c
+++ b/src/dialog.c
@@ -706,7 +706,7 @@ dlg_stop (Dlg_head * h)
     h->state = DLG_CLOSED;
 }

-static void
+void
 refresh_cmd (void)
 {
 #ifdef HAVE_SLANG
diff --git a/src/dialog.h b/src/dialog.h
index 84b51cb..363ce01 100644
--- a/src/dialog.h
+++ b/src/dialog.h
@@ -254,6 +254,7 @@ cb_ret_t default_dlg_callback (Dlg_head * h, Widget * sender, dlg_msg_t msg, int

 /* Default paint routine for dialogs */
 void common_dialog_repaint (Dlg_head *h);
+void refresh_cmd (void);

 #define widget_move(w, _y, _x) tty_gotoyx (((Widget *)(w))->y + _y, ((Widget *)(w))->x + _x)
 #define dlg_move(h, _y, _x) tty_gotoyx (((Dlg_head *)(h))->y + _y, ((Dlg_head *)(h))->x + _x)

[angel_il]

Branch: 2268_FISH_hide_panels_before_connect
changeset: ​https://www.midnight-commander.org/changeset/2268_FISH_hide_panels_before_connect

[andrew_b]

It is not FISH-specific problem. The general solution must be found.

[andrew_b]

Ok, I believe we found a nice solution for FISH. My vote for that.

For other cases (run programs associated with various file types), mc.ext shoult be fixed using 2>/dev/null as Gotar proposed.

[angel_il]

Merged: fa5eaac16236f67e00cfa95701a2a891c24aa382

[zaytsev]

I don't understand why this ticket is closed. The FISH issue is fixed, but there's still a generic issue with the representation of stderr output. mc.ext is not fixed and a solution for stderr corrupting the screen is not implemented.

[andrew_b]

FISH fixups are cherry-picked to 4.7.0-stable branch.
Initial changeset:15f6abf0fc6586afacfd73a1f3e6262faa6664d4