eject -X reports "error while finding CD-ROM name".

Attached find a patch to the original sources which will follow symlinks from the 'device name' line in /proc/sys/dev/cdrom/info.

It assumes that device names listed in the 'drive name' section of /proc/sys/dev/cdrom/info exist in /dev. If this is wrong, then this patch will break portability somewhere.

This allows the -X option to once again work on the current kernel in Hardy.

Debdiff with Adam's patch applied. Could someone in u-m-s look?

Some issues:
 - why does the patch move around the Symlink() function? That's confusing and unnecessary IMHO. If you need it earlier, maybe just add a prototype?
 - calloc/strcat/strcat could be replaced with a single asprintf() to improve readability and robustness
 - resolving symlinks should be done with realpath(3), instead of doing a loop all by yourself
 - can you please report this to upstream?

Updated patch: moved SymLink() back to where it was; added a prototype before the first function that uses it. Used asprintf() instead of calloc/strcat/strcat, and added a configure check for that function. I used a loop to resolve symlinks because that's what the source does in main(), and I figured I'd use the patterns that the original coder used elsewhere. (SymLink() itself does call realpath().) I wanted to change the existing code style as little as possible.

I've already forwarded this (via email; there's no active upstream bugtracker) along with the initial version of the patch. I'm attaching the revised patch, and I'll also send it upstream.

Thanks for the review, and please let me know if you have any other updates or suggestions.

Thanks, Adam. Some more comments:

 - ah, so SymLink() already uses and returns realpath(), so the loop is unnecessary. You'll always get the fully canonicalized file name as a result. This will make the code quite a bit easier.

 - This line

      full_str_name[strlen(full_str_name)-1] = '\0'; // chop trailing \n

   looks fishy. asprintf() does not silently append a line break, so it looks like you'd chop off the last real character? IMHO this like should just go.

 - Indeed you are right wrt. existing code style, the asprintf() configure.in check addition is a bit ugly. But don't worry too much about it, for an Ubuntu patch that's fine. Thanks for sending it upstream, if they know about it, they can change the patch to whichever style they prefer.

Thank you!

The "full_str_name[strlen(full_str_name)-1] = '\0'; // chop trailing \n" is there because str_name was returned with a trailing \n; that is, it looks something like "sr0\n" when it should read "sr0".

Uploaded, now waiting in unapproved queue (will get into intrepid after the beta freeze). Thank you!

This bug was fixed in the package eject - 2.1.5-9ubuntu2

---------------
eject (2.1.5-9ubuntu2) intrepid; urgency=low

  * eject.c: Resolve symbolic links when reading CD-ROM names, to fix "eject
    -X" and possibly other modes. Thanks to Adam Buchbinder for the patch!
    (Patch submitted upstream via email) LP: #264071

 -- Martin Pitt <email address hidden> Mon, 29 Sep 2008 08:20:46 +0200