GNU bug report logs - #25778
25.1; [PATCH] Drastically simplify xdg-open check

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 25778 in the body.
You can then email your comments to 25778 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
To: bug-gnu-emacs <at> gnu.org
Subject: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Fri, 17 Feb 2017 19:16:48 +0100

[Message part 1 (text/plain, inline)]

A user on #emacs brought it to my attention that the xdg-open detection
for browse-url is more complex than it should be and fails detecting
their Linux desktop.  Looking at the definition of
`browse-url-can-use-xdg-open' reveals that it replicates the desktop
environment check as done by `xdg-open` (which is done to detect the
DE-specific default browser setting, not to prevent people not using a
popular desktop environment from using xdg-open).  As enumerating all
possible kinds of Linux desktop is a futile endeavor, I've completely
eliminated this check from it so that the only ones left are whether
we're in a graphical session (by checking $DISPLAY) and whether
`xdg-open` has been found on $PATH.

[0001-Drastically-simplify-check-for-xdg-open.patch (text/x-patch, attachment)]

Message #8 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
Cc: 25778 <at> debbugs.gnu.org
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Fri, 17 Feb 2017 20:38:42 +0200

> From: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
> Date: Fri, 17 Feb 2017 19:16:48 +0100
> 
> A user on #emacs brought it to my attention that the xdg-open detection
> for browse-url is more complex than it should be and fails detecting
> their Linux desktop.  Looking at the definition of
> `browse-url-can-use-xdg-open' reveals that it replicates the desktop
> environment check as done by `xdg-open` (which is done to detect the
> DE-specific default browser setting, not to prevent people not using a
> popular desktop environment from using xdg-open).  As enumerating all
> possible kinds of Linux desktop is a futile endeavor, I've completely
> eliminated this check from it so that the only ones left are whether
> we're in a graphical session (by checking $DISPLAY) and whether
> `xdg-open` has been found on $PATH.

First, checking $DISPLAY doesn't yet mean you are in a GUI frame.  We
have display-graphic-p for that.

And second, I don't understand what will happen with all the bugs
mentioned in the comments if we remove the code that was supposed to
avoid hitting them.  Maybe I'm missing something obvious.

Message #11 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 25778 <at> debbugs.gnu.org
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Fri, 17 Feb 2017 19:55:14 +0100

> First, checking $DISPLAY doesn't yet mean you are in a GUI frame.  We
> have display-graphic-p for that.

The assumption that you need to be in a GUI frame is incorrect.  If I
run `emacs -nw` in a graphical terminal emulator, why would that prevent
me from opening a graphical browser?  The only way I see to detect
whether a graphical browser can be opened is by checking for X with
$DISPLAY.

> And second, I don't understand what will happen with all the bugs
> mentioned in the comments if we remove the code that was supposed to
> avoid hitting them.  Maybe I'm missing something obvious.

OK, I'll go through them all for completeness' sake:

- bug#7166: This one arises from using /bin/sh in combination with
  `nohup` for launching xdg-open, resulting in a lack of shell quoting.
  As `xdg-open` is launched directly, it's no longer relevant and only
  mentioned because of `nohup` (which is no longer used either).
- bug#8917: The old usage of `call-process` has made `nohup` obsolete.
  Irrelevant as `nohup` isn't used any longer.
- bug#9779: Extension of the now removed xdg-open DE detection logic,
  together with some discussion whether nohup is still necessary.  Some
  voices in favor of not doing any unnecessary detection.
- The linked emacs-devel thread: Sighting of a possible bug in
  gnome-open.  As `nohup` is no longer used, there's no way this could
  have been worked around with its usage

Summary: All linked bugs are related to `nohup` which is only checked
for, but not actually used.  Therefore they can be disregarded.

Do you have any actual objections based on a situation where one would
*not* want to use xdg-open, even though it's installed and the user in a
X11 session?

Message #14 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Glenn Morris <rgm <at> gnu.org>
To: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
Cc: 25778 <at> debbugs.gnu.org, Eli Zaretskii <eliz <at> gnu.org>
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Fri, 17 Feb 2017 22:39:51 -0500

The area seems a mess, but:

1) nohup was removed from browse-url-xdg-open in 1c0f1a199ca, with no
explanation and as part of a change that seems unrelated.

2) In https://debbugs.gnu.org/9779#20, the author of that change said

    I note that since 22-Jun-2011 browse-url-xdg-open no longer uses
    nohup.  This is either a mistake on my part (I didn't read the
    comment in browse-url-can-use-xdg-open) or it is indeed
    unnecessary and therefore isn't a requirement for xdg-open.

3) In reply, Jan said in https://debbugs.gnu.org/9779#29

    See discussion at
    http://lists.gnu.org/archive/html/emacs-devel/2009-07/msg00279.html.
    nohup is still needed for xdg-open.

And AFAICS that was the last comment about nohup. No-one ever seems to
have followed up Jan's comment.

Message #17 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Glenn Morris <rgm <at> gnu.org>
To: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
Cc: 25778 <at> debbugs.gnu.org
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Fri, 17 Feb 2017 22:46:03 -0500

PS Maybe you are just saying that since no-one (?) complained since
then, nohup can't have been needed after all. It would be better to test
it, though. It seems difficult to test all the possibilities for
xdg-open.

Ref https://debbugs.gnu.org/25234 

Message #20 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
Cc: 25778 <at> debbugs.gnu.org
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Sat, 18 Feb 2017 09:58:18 +0200

> Date: Fri, 17 Feb 2017 19:55:14 +0100
> From: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
> Cc: 25778 <at> debbugs.gnu.org
> 
> > First, checking $DISPLAY doesn't yet mean you are in a GUI frame.  We
> > have display-graphic-p for that.
> 
> The assumption that you need to be in a GUI frame is incorrect.  If I
> run `emacs -nw` in a graphical terminal emulator, why would that prevent
> me from opening a graphical browser?  The only way I see to detect
> whether a graphical browser can be opened is by checking for X with
> $DISPLAY.

In that case, maybe we shouldn't test $DISPLAY at all?  Why restrict
this to X if xdg-open is already tested for availability?

> Do you have any actual objections based on a situation where one would
> *not* want to use xdg-open, even though it's installed and the user in a
> X11 session?

See Glenn's response about that.  I think we should be more careful
with these issues, and at least do some more research of them before
we can be sure they are no longer relevant.  If you could follow up on
that, it would be great.

Thanks.

Message #23 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 25778 <at> debbugs.gnu.org
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Sat, 18 Feb 2017 09:50:25 +0100

> In that case, maybe we shouldn't test $DISPLAY at all?  Why restrict
> this to X if xdg-open is already tested for availability?

To quote its man page: "xdg-open is for use inside a desktop session
only.".  If you attempt using it outside one (like, on the Linux
console), M-x browse-url fails silently.

Message #26 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
To: Glenn Morris <rgm <at> gnu.org>
Cc: 25778 <at> debbugs.gnu.org
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Sat, 18 Feb 2017 09:55:29 +0100

> PS Maybe you are just saying that since no-one (?) complained since
> then, nohup can't have been needed after all.

Pretty much.  Note that `gnome-open` belongs to GNOME 2 which last
release has been in 2010.  Only Debian Squeeze is offering it in its
package archives and as that one has been EOL'd, I don't see any reason
to pursue it any further.  If `gvfs-open` exhibits the same behavior,
maybe, but I still consider it upstream's fault if they have weird open
behavior, not an area of Emacs to clumsily work around.

Message #29 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Glenn Morris <rgm <at> gnu.org>
To: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
Cc: 25778 <at> debbugs.gnu.org
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Sat, 18 Feb 2017 19:31:46 -0500

Vasilij Schneidermann wrote:

> If `gvfs-open` exhibits the same behavior,

It does. https://bugzilla.gnome.org/show_bug.cgi?id=652262
And https://debbugs.gnu.org/25234, as previously mentioned.

> maybe, but I still consider it upstream's fault

As you can see from that above report, they don't show any signs of
fixing it.

> if they have weird open behavior, not an area of Emacs to clumsily
> work around.

The people who implemented support for xdg-open in Emacs disagreed.

Message #32 received at 25778 <at> debbugs.gnu.org (full text, mbox):

From: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
To: Glenn Morris <rgm <at> gnu.org>
Cc: 25778 <at> debbugs.gnu.org
Subject: Re: bug#25778: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Sun, 19 Feb 2017 14:06:50 +0100

> As you can see from that above report, they don't show any signs of
> fixing it.

Well, fine.  I'm sorry for not noticing a five year old piece of code
that got mistakenly removed without anyone complaining.  What I don't
understand is why my patch (removing IMHO unnecessary DE detection) is
related to using nohup to work around a problem with xdg-open and
gnome-open/gvfs-open.  If you consider fixing the nohup problem a
prerequisite before this patch can be applied, OK, do that then, I can
wait until that's done and recommence discussion on the other issue.

That aside, I cannot reproduce any weird behavior, be it with or without
nohup and gvfs-open.  Both `(call-process "gvfs-open" nil 0 nil
"foo.pdf")` and `(call-process "nohup" nil 0 nil "gvfs-open"
"foo.pdf")` spawn my PDF reader successfully and don't do anything weird
after closing it.  Same if I pass "http://ix.de/" as last argument.  I
therefore consider the nohup issue a case of cargo-culting and would
like to drop its discussion entirely as it does not help with the
current issue.

Message #37 received at 25778-done <at> debbugs.gnu.org (full text, mbox):

From: Paul Eggert <eggert <at> cs.ucla.edu>
To: Vasilij Schneidermann <v.schneidermann <at> gmail.com>
Cc: 25778-done <at> debbugs.gnu.org
Subject: Re: 25.1; [PATCH] Drastically simplify xdg-open check
Date: Thu, 9 Mar 2017 19:13:18 -0800

[Message part 1 (text/plain, inline)]

Thanks for the bug report. I installed the attached somewhat-more-ambitious 
patch that works along the line that you suggested. Although every heuristic in 
this area will be wrong sometimes, I think nowadays we're probably better off 
simply using xdg-open if it's available and if there's a display.

It turns out that nohup is ineffective on xdg-open, as xdg-open's child does a 
'signal (SIGHUP, SIG_DFL)' before it execs (!). This may help to explain why 
nobody noticed when Emacs stopped using nohup. Although nohup may have been 
needed for ancient GNOME versions, I don't think we need to worry about these 
old GNOME versions in future Emacs releases.

I'm closing the bug report as I think the bug is fixed. We can reopen it if I'm 
wrong.

[0001-Simplify-checks-for-xdg-open-and-xdg-email.patch (text/x-diff, attachment)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.