1# @(#)TOUR 8.1 (Berkeley) 5/31/93 2# $FreeBSD$ 3 4NOTE -- This is the original TOUR paper distributed with ash and 5does not represent the current state of the shell. It is provided anyway 6since it provides helpful information for how the shell is structured, 7but be warned that things have changed -- the current shell is 8still under development. 9 10================================================================ 11 12 A Tour through Ash 13 14 Copyright 1989 by Kenneth Almquist. 15 16 17DIRECTORIES: The subdirectory bltin contains commands which can 18be compiled stand-alone. The rest of the source is in the main 19ash directory. 20 21SOURCE CODE GENERATORS: Files whose names begin with "mk" are 22programs that generate source code. A complete list of these 23programs is: 24 25 program input files generates 26 ------- ----------- --------- 27 mkbuiltins builtins builtins.h builtins.c 28 mkinit *.c init.c 29 mknodes nodetypes nodes.h nodes.c 30 mksyntax - syntax.h syntax.c 31 mktokens - token.h 32 33There are undoubtedly too many of these. Mkinit searches all the 34C source files for entries looking like: 35 36 RESET { 37 x = 2; /* executed when the shell does a longjmp 38 back to the main command loop */ 39 } 40 41It pulls this code out into routines which are when particular 42events occur. The intent is to improve modularity by isolating 43the information about which modules need to be explicitly 44initialized/reset within the modules themselves. 45 46Mkinit recognizes several constructs for placing declarations in 47the init.c file. 48 INCLUDE "file.h" 49includes a file. The storage class MKINIT makes a declaration 50available in the init.c file, for example: 51 MKINIT int funcnest; /* depth of function calls */ 52MKINIT alone on a line introduces a structure or union declara- 53tion: 54 MKINIT 55 struct redirtab { 56 short renamed[10]; 57 }; 58Preprocessor #define statements are copied to init.c without any 59special action to request this. 60 61EXCEPTIONS: Code for dealing with exceptions appears in 62exceptions.c. The C language doesn't include exception handling, 63so I implement it using setjmp and longjmp. The global variable 64exception contains the type of exception. EXERROR is raised by 65calling error. EXINT is an interrupt. 66 67INTERRUPTS: In an interactive shell, an interrupt will cause an 68EXINT exception to return to the main command loop. (Exception: 69EXINT is not raised if the user traps interrupts using the trap 70command.) The INTOFF and INTON macros (defined in exception.h) 71provide uninterruptible critical sections. Between the execution 72of INTOFF and the execution of INTON, interrupt signals will be 73held for later delivery. INTOFF and INTON can be nested. 74 75MEMALLOC.C: Memalloc.c defines versions of malloc and realloc 76which call error when there is no memory left. It also defines a 77stack oriented memory allocation scheme. Allocating off a stack 78is probably more efficient than allocation using malloc, but the 79big advantage is that when an exception occurs all we have to do 80to free up the memory in use at the time of the exception is to 81restore the stack pointer. The stack is implemented using a 82linked list of blocks. 83 84STPUTC: If the stack were contiguous, it would be easy to store 85strings on the stack without knowing in advance how long the 86string was going to be: 87 p = stackptr; 88 *p++ = c; /* repeated as many times as needed */ 89 stackptr = p; 90The following three macros (defined in memalloc.h) perform these 91operations, but grow the stack if you run off the end: 92 STARTSTACKSTR(p); 93 STPUTC(c, p); /* repeated as many times as needed */ 94 grabstackstr(p); 95 96We now start a top-down look at the code: 97 98MAIN.C: The main routine performs some initialization, executes 99the user's profile if necessary, and calls cmdloop. Cmdloop 100repeatedly parses and executes commands. 101 102OPTIONS.C: This file contains the option processing code. It is 103called from main to parse the shell arguments when the shell is 104invoked, and it also contains the set builtin. The -i and -m op- 105tions (the latter turns on job control) require changes in signal 106handling. The routines setjobctl (in jobs.c) and setinteractive 107(in trap.c) are called to handle changes to these options. 108 109PARSING: The parser code is all in parser.c. A recursive des- 110cent parser is used. Syntax tables (generated by mksyntax) are 111used to classify characters during lexical analysis. There are 112four tables: one for normal use, one for use when inside single 113quotes and dollar single quotes, one for use when inside double 114quotes and one for use in arithmetic. The tables are machine 115dependent because they are indexed by character variables and 116the range of a char varies from machine to machine. 117 118PARSE OUTPUT: The output of the parser consists of a tree of 119nodes. The various types of nodes are defined in the file node- 120types. 121 122Nodes of type NARG are used to represent both words and the con- 123tents of here documents. An early version of ash kept the con- 124tents of here documents in temporary files, but keeping here do- 125cuments in memory typically results in significantly better per- 126formance. It would have been nice to make it an option to use 127temporary files for here documents, for the benefit of small 128machines, but the code to keep track of when to delete the tem- 129porary files was complex and I never fixed all the bugs in it. 130(AT&T has been maintaining the Bourne shell for more than ten 131years, and to the best of my knowledge they still haven't gotten 132it to handle temporary files correctly in obscure cases.) 133 134The text field of a NARG structure points to the text of the 135word. The text consists of ordinary characters and a number of 136special codes defined in parser.h. The special codes are: 137 138 CTLVAR Variable substitution 139 CTLENDVAR End of variable substitution 140 CTLBACKQ Command substitution 141 CTLBACKQ|CTLQUOTE Command substitution inside double quotes 142 CTLESC Escape next character 143 144A variable substitution contains the following elements: 145 146 CTLVAR type name '=' [ alternative-text CTLENDVAR ] 147 148The type field is a single character specifying the type of sub- 149stitution. The possible types are: 150 151 VSNORMAL $var 152 VSMINUS ${var-text} 153 VSMINUS|VSNUL ${var:-text} 154 VSPLUS ${var+text} 155 VSPLUS|VSNUL ${var:+text} 156 VSQUESTION ${var?text} 157 VSQUESTION|VSNUL ${var:?text} 158 VSASSIGN ${var=text} 159 VSASSIGN|VSNUL ${var:=text} 160 161In addition, the type field will have the VSQUOTE flag set if the 162variable is enclosed in double quotes. The name of the variable 163comes next, terminated by an equals sign. If the type is not 164VSNORMAL, then the text field in the substitution follows, ter- 165minated by a CTLENDVAR byte. 166 167Commands in back quotes are parsed and stored in a linked list. 168The locations of these commands in the string are indicated by 169CTLBACKQ and CTLBACKQ+CTLQUOTE characters, depending upon whether 170the back quotes were enclosed in double quotes. 171 172The character CTLESC escapes the next character, so that in case 173any of the CTL characters mentioned above appear in the input, 174they can be passed through transparently. CTLESC is also used to 175escape '*', '?', '[', and '!' characters which were quoted by the 176user and thus should not be used for file name generation. 177 178CTLESC characters have proved to be particularly tricky to get 179right. In the case of here documents which are not subject to 180variable and command substitution, the parser doesn't insert any 181CTLESC characters to begin with (so the contents of the text 182field can be written without any processing). Other here docu- 183ments, and words which are not subject to splitting and file name 184generation, have the CTLESC characters removed during the vari- 185able and command substitution phase. Words which are subject to 186splitting and file name generation have the CTLESC characters re- 187moved as part of the file name phase. 188 189EXECUTION: Command execution is handled by the following files: 190 eval.c The top level routines. 191 redir.c Code to handle redirection of input and output. 192 jobs.c Code to handle forking, waiting, and job control. 193 exec.c Code to do path searches and the actual exec sys call. 194 expand.c Code to evaluate arguments. 195 var.c Maintains the variable symbol table. Called from expand.c. 196 197EVAL.C: Evaltree recursively executes a parse tree. The exit 198status is returned in the global variable exitstatus. The alter- 199native entry evalbackcmd is called to evaluate commands in back 200quotes. It saves the result in memory if the command is a buil- 201tin; otherwise it forks off a child to execute the command and 202connects the standard output of the child to a pipe. 203 204JOBS.C: To create a process, you call makejob to return a job 205structure, and then call forkshell (passing the job structure as 206an argument) to create the process. Waitforjob waits for a job 207to complete. These routines take care of process groups if job 208control is defined. 209 210REDIR.C: Ash allows file descriptors to be redirected and then 211restored without forking off a child process. This is accom- 212plished by duplicating the original file descriptors. The redir- 213tab structure records where the file descriptors have been dupli- 214cated to. 215 216EXEC.C: The routine find_command locates a command, and enters 217the command in the hash table if it is not already there. The 218third argument specifies whether it is to print an error message 219if the command is not found. (When a pipeline is set up, 220find_command is called for all the commands in the pipeline be- 221fore any forking is done, so to get the commands into the hash 222table of the parent process. But to make command hashing as 223transparent as possible, we silently ignore errors at that point 224and only print error messages if the command cannot be found 225later.) 226 227The routine shellexec is the interface to the exec system call. 228 229EXPAND.C: Arguments are processed in three passes. The first 230(performed by the routine argstr) performs variable and command 231substitution. The second (ifsbreakup) performs word splitting 232and the third (expandmeta) performs file name generation. 233 234VAR.C: Variables are stored in a hash table. Probably we should 235switch to extensible hashing. The variable name is stored in the 236same string as the value (using the format "name=value") so that 237no string copying is needed to create the environment of a com- 238mand. Variables which the shell references internally are preal- 239located so that the shell can reference the values of these vari- 240ables without doing a lookup. 241 242When a program is run, the code in eval.c sticks any environment 243variables which precede the command (as in "PATH=xxx command") in 244the variable table as the simplest way to strip duplicates, and 245then calls "environment" to get the value of the environment. 246 247BUILTIN COMMANDS: The procedures for handling these are scat- 248tered throughout the code, depending on which location appears 249most appropriate. They can be recognized because their names al- 250ways end in "cmd". The mapping from names to procedures is 251specified in the file builtins, which is processed by the mkbuilt- 252ins command. 253 254A builtin command is invoked with argc and argv set up like a 255normal program. A builtin command is allowed to overwrite its 256arguments. Builtin routines can call nextopt to do option pars- 257ing. This is kind of like getopt, but you don't pass argc and 258argv to it. Builtin routines can also call error. This routine 259normally terminates the shell (or returns to the main command 260loop if the shell is interactive), but when called from a builtin 261command it causes the builtin command to terminate with an exit 262status of 2. 263 264The directory bltins contains commands which can be compiled in- 265dependently but can also be built into the shell for efficiency 266reasons. The makefile in this directory compiles these programs 267in the normal fashion (so that they can be run regardless of 268whether the invoker is ash), but also creates a library named 269bltinlib.a which can be linked with ash. The header file bltin.h 270takes care of most of the differences between the ash and the 271stand-alone environment. The user should call the main routine 272"main", and #define main to be the name of the routine to use 273when the program is linked into ash. This #define should appear 274before bltin.h is included; bltin.h will #undef main if the pro- 275gram is to be compiled stand-alone. 276 277CD.C: This file defines the cd and pwd builtins. 278 279SIGNALS: Trap.c implements the trap command. The routine set- 280signal figures out what action should be taken when a signal is 281received and invokes the signal system call to set the signal ac- 282tion appropriately. When a signal that a user has set a trap for 283is caught, the routine "onsig" sets a flag. The routine dotrap 284is called at appropriate points to actually handle the signal. 285When an interrupt is caught and no trap has been set for that 286signal, the routine "onint" in error.c is called. 287 288OUTPUT: Ash uses it's own output routines. There are three out- 289put structures allocated. "Output" represents the standard out- 290put, "errout" the standard error, and "memout" contains output 291which is to be stored in memory. This last is used when a buil- 292tin command appears in backquotes, to allow its output to be col- 293lected without doing any I/O through the UNIX operating system. 294The variables out1 and out2 normally point to output and errout, 295respectively, but they are set to point to memout when appropri- 296ate inside backquotes. 297 298INPUT: The basic input routine is pgetc, which reads from the 299current input file. There is a stack of input files; the current 300input file is the top file on this stack. The code allows the 301input to come from a string rather than a file. (This is for the 302-c option and the "." and eval builtin commands.) The global 303variable plinno is saved and restored when files are pushed and 304popped from the stack. The parser routines store the number of 305the current line in this variable. 306 307DEBUGGING: If DEBUG is defined in shell.h, then the shell will 308write debugging information to the file $HOME/trace. Most of 309this is done using the TRACE macro, which takes a set of printf 310arguments inside two sets of parenthesis. Example: 311"TRACE(("n=%d0, n))". The double parenthesis are necessary be- 312cause the preprocessor can't handle functions with a variable 313number of arguments. Defining DEBUG also causes the shell to 314generate a core dump if it is sent a quit signal. The tracing 315code is in show.c. 316