update error in Debian 8 Linux

[SilverBear]

Hi, all.
And @ Kovid: Hope your vacation was wonderful!

I've encountered the same error message twice --a few weeks ago when trying to update from Calibre 2.46 to 2.47 --and today, trying to update 2.46 to 2.48.

I run the DL/install command --cut & paste from the calibre website. The download completes, and as the installation starts to run, it terminates with these messages:

. . .
Downloaded 62233416 bytes
Checking downloaded file integrity...
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "<string>", line 670, in main
File "<string>", line 655, in run_installer
File "<string>", line 631, in download_and_extract
File "/usr/lib/python2.7/shutil.py", line 232, in rmtree
onerror(os.path.islink, path, sys.exc_info())
File "/usr/lib/python2.7/shutil.py", line 230, in rmtree
raise OSError("Cannot call rmtree on a symbolic link")
OSError: Cannot call rmtree on a symbolic link

. . . and the process terminates.

I've been running Calibre on this desktop machine under a few different Linux distros over the last four years with no trouble. I've had Debian 8 ("Jessie") installed & calibre has been happy since June 2015. Does anyone have advice on how to work around this problem?

[eschwartz]

Are there symbolic links in /opt/calibre ?
There shouldn't be, unless something other than calibre's installer has touched that folder...

But anyway, try clearing it with `sudo rm -rf /opt/calibre` and retrying the installation.

Shameless plug: Also see my sticky in Related Tools, for an easy way to auto-update.

[kovidgoyal]

That's a very strange error, as far as I can tell it should be impossible, unless some other process is modifying the installation directory simultaneously with the calibre installer. Or the lstat system call is failing for files in the installation directory.

@eschwartz: The presence of symlinks should not matter, since python's rmtree implementation does not call itself with symlinks as arguments. The only way it can happen is if os.lstat fails for a path. Then the rmtree() implementation just assumes the path is a directory. I dont know why rmtree was designed to ignore lstat failures.

[SilverBear]

Originally Posted by kovidgoyal

That's a very strange error...

Yes, it is. And it turns out that I am the very strange man who caused it.

In addition to my Debian 8 install, I've been testing out the latest Linux Mint and MX-Linux versions. I routinely symlink my data partitions into my /home/silverbear directories so I have access to all my data no matter what distro I'm running.

APPARENTLY I got the not-so-bright idea to do the same with my install of calibre 2.46 as a time-saving experiment. . .
. . . and then forgot I'd done it.

Lesson: do not modify the OS late at night when you have an insomnia attack!

I'm sorry for wasting your time, gentlemen. Symlink has been rm'd and calibre 2.48 is nicely installed in /opt with no problems.

[eschwartz]

Originally Posted by kovidgoyal

@eschwartz: The presence of symlinks should not matter, since python's rmtree implementation does not call itself with symlinks as arguments. The only way it can happen is if os.lstat fails for a path. Then the rmtree() implementation just assumes the path is a directory. I dont know why rmtree was designed to ignore lstat failures.

Sorry presence of symlinks ==> is a symlink (if I read the docs right).
It is 2 AM here (was: midnight) -- this is probably a sign that I should go to sleep now...

@SilverBear -- not just you...

[kovidgoyal]

This should take care of that case:

https://github.com/kovidgoyal/calibr...e70cd0321e2f85