Program 2 • Function pointers, a shell Solution




$Id:,v 1.8 2017-07-05 15:01:12-07 – – $


PWD: /afs/




  1. Overview


You will maintain a tree structure with a simple hierarchy, maintained by a map of functions. Programming will be done C++ style, not C style, as shown in the follow-ing table.


Do not use : Instead, use :
char* strings <string>
C arrays <vector>
<stdio.h> <iostream>, <iomanip>
pointers <shared_ptr>
union inheritance
<header.h> <cheader>


Header files : Include only C++11 header files and facilities where feasable, and use namespace std. Include <cheader> files only when C++-style files are unavailable. Include <header.h> files from C only when an appropriate <cheader> files is unavailable. Use the script (a wrapper for to check style.



  1. Program Specification


The program specification is given in terms of a Unix man(1) page.




yshell — in memory simulated tree shell



yshell [-@ flags]




This shell reads commands from stdin and write output to stdout, with errors being set to stderr. Each line read by the shell is parsed into words by split-ting using space characters, with any number of spaces between words. There may also be leading and trailing spaces. The first word on any line is a com-mand to be simulated, and the rest are operands to that command. If either stdin or stdout is not a tty, each line from stdin is echoed to stdout.



The commands modify an inode tree, where each inode is either a file or a directory. Files contain data and directories contain inodes. An inode is speci-fied by means of a pathname. A pathname consists of a sequence of characters separated by slash (/) characters.


The inode tree has a root, which is a special node, and also a current inode as well. Whenever a pathname is decoded, if the first character is a slash (/), decoding begins at the root, otherwise it begins with the current directory. Whenever a pathname component is a dot (.), it refers to the current directory. If a component is a double dot (..) it refers to the parent of the current direc-tory. Every directory has both of these entries, with the root being its own par-ent. Multiple adjacent slashes are treated as a single slash. Trailing slashes


are permitted only on directories.


Every inode has three attributes : an inode number, which is uniquely assigned, starting from 1 for the root ; contents, which is a map from file-names to inodes for a directory, and text for a file ; and a size, which is the byte count for text, and the number of sub-inodes for a directory.




None. All input comes from stdin.




The -@ option is followed by a sequence of flags to enable debug output, written to stderr.




The following commands are interpreted. Error messages are printed and nothing is done in the case of invalid operands.


  • string


If the first non-space character on a line is a hash, the line is a comment and is ignored.


cat pathname . . .


The contents of each file is copied to stdout. An error is reported if no files are specified, a file does not exist, or is a directory.


cd [pathname]


The current directory is set the the pathname given. If no pathname is specified, the root directory (/) is used. It is an error if the pathname does not exist or is a plain file, or if more than one operand is given.


echo [words . . .]


The string, which may be empty, is echoed to stdout on a line by itself.


exit [status]


Exit the program with the given status. If the status is missing, exit with status 0. If a non-numeric argument is given, exit with status 127.


ls [pathname . . .]


A description of the files or directories are printed to stdout. It is an error if any of the file or directory does not exist. If no pathname is speci-fied, the current working directory is used. If a pathname specified is a directory, then the contents of the directory are listed. A directory listed within a directory is shown by a terminating slash. Elements of a direc-tory are listed lexicographically.



For each file listed, output consists of the inode number, then the size, then the filename. Output is lined up into columns and each column is separated from the next by two spaces. The numeric fields are exactly 6 characters wide and the units position in a column must be aligned.


lsr [pathname . . .]


As for ls, but a recursive depth-first preorder traversal is done for subdi-rectories.


make pathname [words . . .]


The file specified is created and the rest of the words are put in that file. If the file already exists, a new one is not created, but its contents are replaced. It is an error to specify a directory. If there are no words, the file is empty.


mkdir pathname


A new directory is created. It is an error if a file or directory of the same name already exists, or if the complete pathname to the parent of this new directory does not already exist. Two entries are added to the direc-tory, namely dot (.) and dotdot (..). Directory entries are always kept in sorted lexicographic order.


prompt string


Set the prompt to the words specified on the command line. Each word is separated from the next by one space and the prompt itself is terminated by an extra space. The default prompt is a single percent sign and a space (% ).




Prints the current working directory.


rm pathname


The specified file or directory is deleted (removed from its parent’s list of files and subdirectories). It is an error for the pathname not to exist. If the pathname is a directory, it must be empty.


rmr pathname


A recursive removal is done, using a depth-first postorder traversal.




  • No errors were detected.


  • Error messages were printed to cerr.


  1. A Sample Run


The following table shows a sample run. Each interaction with the shell is listed in a separate box with shell output in Courier Roman and user input in Courier Bold typeface. A commentary about what is happening is opposite in the right column.


% pwd   Initially the cwd is the root directory.
% ls   The absence of an operand to ls means
/:   that dot is used as its operand, which
1 2  . is  currently  the  root.   Directories
1 2  .. always  contain  at  least  two  items,
    namely  dot  and  dotdot.  The  inode
    number of the root is always inode #1.
    The parent of dotdot is itself.


  4 of 7  
% make foo this is a test Make a file called foo which contains
      the string ‘‘this is a test’’, which is
      14 characters.  An inode is allocated,
      namely inode #2.  
% make bar test a is this Another  file,  similarly  created,  with
      inode #3.  
% ls     Same  as  the  previous  output  of  ls,
/:     except with two more files.  Note that
1 4  . files are kept in lexicographic order, so
1 4  .. bar is listed before foo.  
3 14 bar    
2 14 foo    
% cat food     An error message is printed, causing
cat: food: No such file or directory the return code from the shell eventu-
      ally to be 1 rather than 0.  Note the
      error  format :  command  followed  by
      object causing the problem followed by
      the reason for the failure.  
% cat foo     Files  can  consist  of  only  one  line,
this is a test   namely a string.  
% echo O for a muse of fire Arguments to echo are simply written
O for a muse of fire to the standard output.  
% prompt =>     The prompt is changed to the charac-
      ters ‘‘=>’’ followed by a space. Multiple
      words would have been permitted.
=> rm bar     The file bar is deleted and the size of
      the root directory is reduced by 1.
=> make baz foo bar baz A new file is created with inode #4.
=> mkdir test   Inode  #5  is  created  as  a  directory
      called test. This directory is a child of
      the root and contains the two usual
      entries, dot and dotdot.  
=> prompt %     The prompt is changed back to a % fol-
      lowed by a space.  
% ls /     Just checking the contents of the root.
1 5  .    
1 5  ..    
4 11 baz    
2 14 foo    
5 2 test/    
% cd test     The cwd is now test.  
% pwd     Yes, it is.  


  5 of 7  
% cd     Without arguments cd goes back to the
      root directory.  
% pwd     OK.  
% cd test     Go to a directory called test which is a
      subdirectory of the cwd, whose alias
      name is always dot.  
% pwd        
% cd ..     Dotdot is always an alias for the par-
      ent of the cwd.  
% pwd        
% cd test     This would have errored out if test
% make me me me me were not a directory or did not exist.
      The next available inode is #6.  
% cat me        
me me me        
% cd ..        
% cd test        
% cat me        
me me me        
% cd        
% lsr /     Recursive  directory  listing. This  is
/:     done using a preorder traversal. With-
1 5 . ing a given level, lexicographic order-
1 5 .. ing is used.  Recursion will go through
4 11 baz all subdirectories at all levels.  
2 14 foo    
5 3 test/    
5 3  .    
1 5  ..    
6 8 me    
% cd test        
% mkdir foo     Note that foo uses inode #7, bar uses
% cd foo     inode #8, and baz uses inode #9.
% mkdir bar        
% cd bar        
% mkdir baz        
% cd baz        
% ls .     At this point dot is baz and dotdot is
.:     bar.  
9 2  .    
8 3  ..    


  6 of 7  
% cd /     A rather large test showing inode num-
% lsr test     bers, file and directory sizes, and file-
/test:     names.  Note that directory names are
5 4 . indicated in the listing with a trailing
1 5 .. slash.  Again, the size of a file is the
7 3 foo/ number of characters in it and the size
6 8 me of a directory is the sum of the number
/test/foo:     of files The subdirectory count is not
7 3 . recursive.  
5 4  ..    
8 3 bar/    
8 3  .    
7 3  ..    
9 2 baz/    
9 2  .    
8 3  ..    
% ^D     End  of  file  or  Control/D  causes  the
      shell to exit.  


  1. A Tour of the Code


Begin by studying the code provided in the code/ subdirectory. There are four mod-ules arranged into a header (.h) file and an implementation (.cpp) file, the main pro-gram in main.cpp, and, of course, a Makefile. Notice that comments are in the header for when specifying general functionality, and only in the implementation as a way of explaining how something works.




Study the various targets all, ${EXECBIN}, %.o, ci, lis, clean, spotless, submit, verify, and deps, which perform their usual functions.




The debug module is already written for you. It is useful in tracing through your code. Other parts of the code may want to have more DEBUGF and DEBUGS calls added. Note that you should also use gdb to track down bugs. Use valgrind to check for invalid memory references and memory leak.




The util module is just a collection of independent functions which are herded together. It is a module without any cohe-sion, but a useful place to park various random functions.




The main program is mostly complete.


loops reading and executing commands.
It scans options, then




Note that the functions are provided but do not do more than print a trace. Each function as a inode_state argument, passed by reference, which it might update, and a wordvec arl-gument. words[0] is the name of the command itself, so the first argument is words[1]. This will take the most work, but commands can be added one at a time, by addition to the


7 of 7


organization that is already there. You may add private func-tions is you need to. This is the major execution engine.


file_sys.{h,cpp}           The inode, or file system, module is the main data structure that you are working on. As you implement commands, also implement the functions of this module as well.


  1. What to Submit


Submit the files Makefile, README, and all C++ header and source files. All header files must end with .h, and source files must end with either .cpp or .cc as the suf-fix. You must be consistent and use one or the other suffix for all files.


Run gmake to verify that the build is possible. And when you run submit, do so from the Makefile target of that name. That way you won’t forget to submit a file. If you forget to submit a file, or submit the wrong version, you will lose at least 50% of your program’s value. Make sure that you do not get any warnings from g++ and that checksource does not complain about anything. Do not submit any file that is built by the Makefile.


If you are doing pair programming, follow the additional instructions in :




The code must compile and run using g++ on, regardless of whether it runs elsewhere. As of the time of formatting this document, that was :


bash-1$ g++ –version | head -1


g++ (GCC) 5.3.1 20160406 (Red Hat 5.3.1-6)


bash-2$ uname -srp


Linux 3.10.0-514.16.1.el7.x86_64 x86_64 bash-3$ hostname


Compile your code with


g++ -std=gnu++14 -g -O0 -Wall -Wextra -Wold-style-cast

error: Content is protected !!