1

I have a scenario where I am trying to find multiple files based on the output of another command.

The command itself is

./libopt -d test . 2>&1 | sed -rn 's/Cannot find library \"(.*?)\"/\1/p'

and this gives me the following output

libopcodes-2.25-system.so
libbfd-2.25-system.so
libz.so.1
libdl.so.2
libc.so.6
libosal.so
libarchive.so.13
libcrypto.so.1.0.0
libfreetype.so.6
libpng12.so.0
libpthread.so.0
libts-1.0.so.0
libgcc_s.so.1
libssl.so.1.0.0
libm.so.6
ld-linux.so.3
libxml2.so.2
libbz2.so.1.0
liblzma.so.5
liblzo2.so.2
libattr.so.1
libacl.so.1
libnettle.so.4

Is there anyway that I can feed this input to the following find command in order to find the location of each of these files?

find /usr -name <INPUT_LINE>
Improve this question
1

3 Answers 3

Reset to default
2

On the assumption that no problematic characters (e.g. those in IFS) exist in the library names, loop over the list and pass each to a new find:

./libopt -d test . 2>&1 | sed -rn 's/Cannot find library \"(.*?)\"/\1/p' \
| while read libname; do
  find /usr -name "$libname"
done

Otherwise, for a single filesystem pass (the above does one for each input file), the find -name option is not suitable for matching a list of filenames, so perhaps the output of a single filesystem search could be passed to grep. The search also could be narrowed by only looking in directories under /usr that are named lib:

./libopt -d test . 2>&1 | sed -rn 's/Cannot find library \"(.*?)\"/\1/p' > blah
find /usr/lib /usr/*/lib -type f | grep -F -f blah

Use a mktemp generated filename if blah is a problem. With a fancy shell, the file of patterns for grep could be generated on the fly via some form of process substitution.

Improve this answer
2
  • if using find | grep you could save ./libopt output to a file and use it with find .. | grep -f filename. should probably use -F for fixed strings too. this is much simpler than my method of constructing a long regexp first. Commented Mar 7, 2016 at 21:08
  • @cas yeah, grep -f file I something I keep forgetting about. Commented Mar 8, 2016 at 1:02
0

You could transform the filenames into a single regular expression and use find's -regex or -iregex (case-insensitive) options.

libs=$(./libopt -d test . 2>&1 | 
       sed -rn 's/Cannot find library \"(.*?)\"/\1/p' | 
       xargs -r | sed -e 's/ /|/g ; s/\./\\./g')

[ -n "$libs" ] && find /usr -regextype egrep -iregex ".*/($libs)"

xargs puts all the filenames onto one line separated by spaces, and the following sed changes the spaces to | and . to \.. The find command is only run if $libs is non-empty.

Note: if too many (thousands) of filenames are output by ./libopt, the xargs command will output 2 or more lines, $libs will be too large to fit on one command line, and this will break. If that is likely (it seems unlikely to me, given what ./libopt is outputting), you could work around it by running the xargs | sed and find in a loop, for example:

libs=$(./libopt -d test . 2>&1 | sed -rn 's/Cannot find library \"(.*?)\"/\1/p')

for re in $(echo "$libs" | xargs | sed -e 's/ /|/g ; s/\./\\./g') ; do 
    find . -regextype egrep -iregex ".*/($re)"
done
Improve this answer
0

With zsh:

files=(
  ${(f)"$(./libopt -d test . 2>&1 |
          sed -rn 's/Cannot find library \"(.*?)\"/\1/p')"})
glob=('$files['{1..$#files}']')
eval "paths=(**/(${(j:|:)glob})(D.))"
ls -ld -- $paths

See this answer to a similar question: https://unix.stackexchange.com/a/265815 for an explanation and also some alternative approach if you don't have zsh.

Improve this answer

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.