Ygtjki

I have some weird warnings when zipping a directory's content. I use a command line I use for a ton of websites backup since years and never had such issue, and I try to figure out why I have the warnings (even though the zip files seem created correctly) and I fail to figure it out in a reasonable time (I have put what I found, keep reading). Hopefully someone can help.

I go in the directory I want to backup to a zip file:

cd /home/mickoz/michael.muryn.name

Here is the command I run (I tailored it to debug only and reproduce problem):

zip -r -qv test.zip * .* -x "..*" -x test.log -x test.zip > test.log

For some reason it logs these warnings right at the beginning:

zip warning: name not matched: ../logs/michael.muryn.name/http.2810216.bak/html/daily.html
zip warning: name not matched: ../logs/michael.muryn.name/http.2810216.bak/html/index.html
zip warning: name not matched: ../logs/michael.muryn.name/http.2810216.bak/html/longterm.html
zip warning: name not matched: ../logs/michael.muryn.name/http.2810216.bak/html/monthly.html
zip warning: name not matched: ../logs/michael.muryn.name/http.2810216.bak/html/resources
zip warning: name not matched: ../logs/mickoz.dreamhosters.com/http.2150913.bak/html/index.html
zip warning: name not matched: ../logs/mickoz.dreamhosters.com/http.2150913.bak/html/monthly.html
zip warning: name not matched: ../logs/mickoz.dreamhosters.com/http.2150913.bak/html/resources
zip warning: name not matched: ../logs/muryn.name/http.2810083.bak/html/index.html
zip warning: name not matched: ../logs/muryn.name/http.2810083.bak/html/monthly.html
zip warning: name not matched: ../logs/muryn.name/http.2810083.bak/html/resources
zip warning: name not matched: ../muryn.name/.#svn-commit.tmp

Now, there is NO symlinks in the directory I zip content of (I did execute a command to find if there was even deepdown and there is none).

Heck, just to be sure, I just even tried to create a new directory /home/mickoz/michael.muryn.name2, created two text files in it only. Run the same zip command I put above in that new directory... and the exact same warnings appear! Seem like it tries to search outside of the directory, but I fail to see why as I exclude "..*" in my zip command and that should do the job.

I even tried to do it in a directory named /home/mickoz/testzipbackup in case the dots in the directory name was causing problem and the same result.

Any idea what may be causing this? How can I debug this?

Additional infos:

There was a symbolic link located there

/home/mickoz/muryn.name/.#svn-commit.tmp -> mickoz@here.22973

Take note that the symlink was pointing to something non-existing. This seems important as I debug more and more.

I deleted it and re-executed my test zip command, and this line disappeared from the log:

zip warning: name not matched: ../muryn.name/.#svn-commit.tmp

So it definitely try to search outside the directory for some reason and I fail to see why it does it with what looks like symbolic link only, while excluding files correctly it seems.

From there I dug deeper and did a search for all symbolic links on my user account by using this command in the root of the home directory:

find . -type l -ls

Here is a specific and interesting sample of the result:

14575211295 0 lrwxrwxrwx 1 root root 49 Aug 6 2011 ./logs/muryn.name/http.2810083 -> /home/_domain_logs/mickoz/muryn.name/http.2810083
14145610800 0 lrwxrwxrwx 1 root root 21 Aug 6 2011 ./logs/muryn.name/http.2810083.bak/access.log.0 -> access.log.2011-08-05
14145612338 0 lrwxrwxrwx 1 root root 20 Aug 6 2011 ./logs/muryn.name/http.2810083.bak/error.log.0 -> error.log.2011-08-05
15649147256 0 lrwxrwxrwx 1 root root 67 Aug 5 2011 ./logs/muryn.name/http.2810083.bak/html/daily.html -> /home/mickoz/logs/muryn.name/http.2810083/html/inactive_report.html
15649147257 0 lrwxrwxrwx 1 root root 73 Aug 5 2011 ./logs/muryn.name/http.2810083.bak/html/index.html -> /home/mickoz/logs/muryn.name/http.2810083/html/monthly/2011-08/index.html
15649147258 0 lrwxrwxrwx 1 root root 66 Aug 5 2011 ./logs/muryn.name/http.2810083.bak/html/longterm.html -> /home/mickoz/logs/muryn.name/http.2810083/html/longterm/index.html
15649147259 0 lrwxrwxrwx 1 root root 73 Aug 5 2011 ./logs/muryn.name/http.2810083.bak/html/monthly.html -> /home/mickoz/logs/muryn.name/http.2810083/html/monthly/2011-08/index.html
15649147260 0 lrwxrwxrwx 1 root root 27 Oct 19 2009 ./logs/muryn.name/http.2810083.bak/html/resources -> /home/mickoz/logs/resources
14557793344 0 lrwxrwxrwx 1 root root 49 Mar 19 15:12 ./logs/muryn.name/http -> /home/_domain_logs/mickoz/muryn.name/http.2810083

I analyzed this and found out that only symlinks that are pointing to non-existing target appears in my zip command warnings.

Now, only have to understand why the zip command looks for those symlinks outside of the working directory and only acts/emits a warning on the symlinks pointing to non-existing target.

There are two things going on here.

The first is that

zip -r -qv test.zip * .* -x "..*" -x test.log -x test.zip

is expanded by the shell to list everything matching * and .*, which includes .. in your case.

The second is that zip checks all the file names it’s given, when it’s asked to recurse, before filtering excluded patterns; so it sees ../logs/michael.muryn.name/http.2810216.bak/html/daily.html etc. and calls stat on them, which produces an error, and zip outputs “name not matched” as a result.

The usual way to avoid this is to specify .[^.]* instead of .*, or since you’re also matching *, run

zip -r -qv test.zip . -x test.log -x test.zip

instead.