[Webkit-unassigned] [Bug 224417] New: bwrap: Can't mkdir parents for /run/systemd/resolve/stub-resolv.conf: Read-only file system

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=224417

            Bug ID: 224417
           Summary: bwrap: Can't mkdir parents for
                    /run/systemd/resolve/stub-resolv.conf: Read-only file
                    system
           Product: WebKit
           Version: Other
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: WebKitGTK
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: webkit at livna.org
                CC: bugs-noreply at webkitgtk.org

So I've upgraded from fedora 33 to fedora 34 beta this week, and evolution, which I used on a daily basis so far, refused to start after the upgrade. The process does not crash, it just hangs and the window never pops.

Fedora 34 provides a gnome 40 stack, with evolution-3.40.0-1.fc34.x86_64 and webkit2gtk3-2.32.0-2.fc34.x86_64.

If I try to run 'evolution' from command line, it prints:

bwrap: Can't mkdir parents for /run/systemd/resolve/stub-resolv.conf: Read-only file system

But the evolution window never shows up. I've tried creating a bogus user on my system to see if it was systemd-wide or just about my own user environment and it turned out the bogus user could actually make evolution start

That's when I've started googling for a potential fix and/or workaround. Then I've came accross this year old (and fixed) bug: https://bugzilla.redhat.com/show_bug.cgi?id=1781631 which is linked to https://bugs.webkit.org/show_bug.cgi?id=202880 which looked very similar.

So, I've tried 'WEBKIT_FORCE_SANDBOX=0 evolution': it worked. I can get my e-mails now.

I've read in the bug comments that it was due to an empty component in LD_LIBRARY_PATH.

My own LD_LIBRARY_PATH is non-empty and contains a single component: The current directory '.' (I often drop .so files here and there for debug. It turned out to be quite convenient). The problem occurs whatever the current working directory is, and I had no .so in my $HOME nor the few other directories I've tried, *but* if I 'cd /', then the error message becomes:

bwrap: Can't create file at /.flatpak-info: Read-only file system

which sounded familliar after reading . The same happens if I explicitly `export LD_LIBRARY_PATH=/`.

I know one could consider this bad practice, but it's a use case nonetheless. And as a user, what I want is that LD_LIBRARY_PATH does not interfer with the process I start unless there are actually .so files in the directories pointed by its components.

I have not checked webkitgtk source tree but if there are automated test cases somewhere in your environment, maybe you should have a test case for more LD_LIBRARY_PATH values such has:
1. .
2. ..
3. any relative PATH value
4. /
5. Those values with one or more trailing /
6. Things like that :)

I hope this helps. (and gets fixed :) )

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20210411/f5799adb/attachment.htm>