Re: [hurd-amd64] ibus test failures

[Samuel Thibault]

Hello,

Diego Nieto Cid, le dim. 29 déc. 2024 22:14:40 +0000, a ecrit:
>     (ibus-daemon:17123): GLib-GIO-WARNING **: 20:49:29.230: Expected a 
> credentials
>     struct of 84 bytes but got 88 bytes of data
> 
> which I traced to the GIO function g_unix_credentials_message_deserialize 
> (which
> can be seen here[1]).
> [1] 
> https://gitlab.gnome.org/GNOME/glib/-/blob/main/gio/gunixcredentialsmessage.c?ref_type=heads#L115
> 
> It seems to be some structure size issue on amd64 (i386 tests don't fail) 
> regarding
> SCM_CREDS implementation.

See the error test, it's about G_CREDENTIALS_NATIVE_SIZE, see its
definition:

#define G_CREDENTIALS_NATIVE_SIZE (sizeof (struct cmsgcred))

And the definition of struct cmsgcred in bits/socket.h

I guess on amd64 the struct size somehow gets rounded up to 88 bytes,
but I don't see why.
Possibly it's the position of cmcred_groups that gets aligned, but I
don't see why. At least the sender and the receiver don't seem to agree
on the shape of struct cmsgcred.

> Two patches to the test runner needs to be applied for hurd-amd64[2]:
> 
>   * add t64 suffix to libgtk-3-0 directory where the query immodules binary 
> is found

It's not just on hurd-amd64, it's on all Debian ports. I'm not sure gtk
maintainers are really aware that the directory changed, better discuss
with the Debian maintainer about that.

>   * replace `&> file` redirection to the more standard `> file 2>&1` 
> understood by dash

It's not "on GNU Hurd sh points to dash", but on all Debian systems. And
that should be forwarded to upstream.

Samuel