I have been running emacs (x11 version, unwittingly) on wayland without realizing there is a wayland-specific version of emacs. My desktop has started spontaneously going apeshit and crashing. So my first move was to run emacs this way:

$ emacs --fg-daemon
$ emacsclient -c

The idea was that when the display manager goes to shit, I don’t lose my shit. I should be able to simply reattach to my emacs session by running emacsclient -c. Starting the emacs server yields this msg:

Warning: due to a long standing Gtk+ bug¹ https://gitlab.gnome.org/GNOME/gtk/issues/221 Emacs might crash when run in daemon mode and the X11 connection is unexpectedly lost. Using an Emacs configured with --with-x-toolkit=lucid does not have this problem.

Yikes. Note the keyword might crash. So there is an arbitrary chance of survival. So then I discovered there is a wayland-specific version of emacs (emacs-pgtk). Installing it on debian required removing the original X11 version of emacs and resolving some dependency issues by not installing some pkgs. Fine, I did that. Now when I run emacs-pgtk as a server, it starts with:

Due to a limitation in GTK 3, Emacs built with PGTK will simply exit when a display connection is closed. The problem is especially difficult to fix, such that Emacs on Wayland with multiple displays is unlikely ever to be able to survive disconnects.

Fuck. So now when wayland crashes or is shut down, there is no chance of recovery. WTF? I was perhaps better off running the X11 version on wayland, where there is at least a non-deterministic chance of recovery.

¹ Apparently gitlab.gnome.org is an exclusive access-restricted website, so I am blocked from even reading about the bug.

  • evenwichtOP
    link
    fedilink
    English
    arrow-up
    1
    ·
    8 天前

    The bugs may be old but they don’t seem to be triggered in normal circumstances. Hence why they are not fixed over a long time. The bugs are only killing me now because of other bugs (wayland just recently started going randomly apeshit and crashing for me lately). The server mode is a workaround for those other bugs in my case. But I can see why some advanced users would like to use the server mode anyway.

    Debian offers these 3 versions (possibly more):

    • emacs-gtk
    • emacs-pgtk
    • emacs-lucid

    Installing emacs-lucid after installing emacs-pgtk is a shit show, which then makes emacs-pgtk a terrible default. And as you rightly point out, emacs-gtk is a bad default because it sometimes crashes when the display manager goes down. That leaves emacs-lucid. Perhaps that should be the default on debian when someone runs apt install emacs. Or perhaps even better would be for the installer to interview users to find out which of the 3 is ideal for the user.

    The Debian bug tracker has recently undergone some defensive changes that now block bug submission over tor. So I guess won’t document this idea anywhere that matters.

    • Jerkface (any/all)@lemmy.ca
      link
      fedilink
      English
      arrow-up
      1
      arrow-down
      2
      ·
      8 天前

      Not sure I agree with your premise. It is a perfectly normal use case that happens frequently for everyone using emacs-server, which is intended and expected to outlive interactive sessions like a DE. This wart is a shame on both projects. I suppose you are saying that emacs-server itself is an obscure case, which I can’t really speak to.

      • evenwichtOP
        link
        fedilink
        English
        arrow-up
        1
        ·
        8 天前

        It’s likely rare to use the emacs server mode at all. Most users would consider that extra complexity for nothing as far as their needs go. It has some nuanced advantages for advanced users. Consider that rudel had potential to make server mode broadly interesting for a wide userbase because multiple users could simultaneously collaborate on the same emacs buffer. Rudel is non-debian and not included as part of emacs. And it was last touched 4 years ago IIUC. Then within the already small number of users of the server, some will shutdown the DE and restart it, or switch DEs, or suffer from DE crashes. It’s a quite rare scenario within an uncommonly used feature.

        Of course it’s a clear bug that the server does not outlive the DE, and an embarrassment. The reason it has persisted could only be attributed to lack of demand for a fix. You and I are probably like 2 in a million people who care to have it fixed. I can only speculate since I cannot reach the access-restricted bug tracker on gnome.org. I don’t suppose gitlab has a bug voting feature like launchpad does for ubuntu.