Re: [Gnu-arch-users] [BUG] a socket file in the way causes 'tla inventory' panic

[Jan Hudec]

On Mon, May 03, 2004 at 18:33:45 +0200, Matthieu MOY wrote:
> Quoting Jan Hudec <address@hidden>:
> 
> > Just like for directories and symlinks tla does not try to read the id
> > from the file and instead looks for an explicit one, neither should it
> > with pipes, sockets and devices.
> 
> I'm not sure to understand your answer. It *should* not try to read, but it
> actually does:
> 
> $ strace tla inventory
> [ ... ]
> lstat64("./socket", {st_mode=S_IFSOCK|0755, st_size=0, ...}) = 0
> open("./.arch-ids/socket.id", O_RDONLY) = -1 ENOENT (No such file or 
> directory)
> open("./.arch-ids/=all", O_RDONLY)      = -1 ENOENT (No such file or 
> directory)
> open("./socket", O_RDONLY)              = -1 ENXIO (No such device or address)
> write(2, "error finding file id (", 23error finding file id () = 23
> 
> The last "open" is an attempt to find the tagline, isn't it ?

Sure. I said that is *should* *not*. And I said it does not for
symlinks.

> > Note, however, that socket nodes should
> > never exist unless the application is running, since bind insists on
> > creating them.
> 
> I use an application which lets a socket in my working directory even after
> exiting. It's not my application, but I don't want it to disturb my work 
> anyway.

If the socket name is recognizable by a regular expression, add it to
the junk (or precious or exclude) regexp in =tagging-method as
a quick-fix.

-------------------------------------------------------------------------------
                                                 Jan 'Bulb' Hudec 
<address@hidden>

signature.asc
Description: Digital signature