Compilers

3rd year
Spring term

Mick O’Donnell: [email protected]

Based on the book:

Compiladores e interpretes: teoría y práctica (Estrella Pulido,
Marina de la Cruz, Alfonso Ortega, Manuel Alfonseca)

11 Error Detection

11.1 Introduction

1
 Error Detection
Introduction

Detection of Errors
• To be useful, a compiler should detect all errors in
the source code and report them to the user.
• These errors culd be:
• Lexical errors: e.g., badly formed identifiers or constants, symbols
which are not part of the language, badly formed comments, etc.
• Syntactic errors: chains of syntactic units that do not conform to the
syntax of the source language.
• Semantic errors: e.g., operations conducted on incompatible types,
undeclared variables, double declaration of variable, reference before
assignment, etc.
• Run-time errors: errors detectable solely at run time, pointers with null
value or whose value is outside allowed limits, or indexing of vectors
with unsuitable indices, etc.

Error Detection
Introduction

Detection of Errors
• To be useful, a compiler should detect all errors in
the source code and report them to the user.
• These errors culd be:
• Lexical errors: e.g., badly formed identifiers or constants, symbols which
are not part of the language, badly formed comments, etc.
Compilation• Syntactic errors: chains of syntactic units that do not conform to the syntax
of the source language.
Errors
• Semantic errors: e.g., operations conducted on incompatible types,
undeclared variables, double declaration of variable, reference before
assignment, etc.
• Run-time errors: errors detectable solely at run time, pointers with null
Execution value or whose value is outside allowed limits, or indexing of vectors with
Errors unsuitable indices, etc.

2
 Error Detection
Introduction

A good compiler…
• Reports ALL errors
• Does not falsely report errors.
• Does not repeatly report the same error.

Error Detection
Introduction

Detection of Errors
• How many errors will the c compiler report?
void main () {
int i,j,k;
i=0; /* Asigno 0 a i //
j=;
=k;
}

3
 Error Detection
Introduction

Detection of Errors
• How many errors will the c compiler report?
void main () {
int i,j,k;
i=0; /* Asigno 0 a i //
j=;
=k;
}
• Answer: 2
1) Comment on line 3 not terminated
2) Function main has no closing bracket
7

Error Detection
Introduction

False Detection of Errors

• Sometimes, due to one error, the compiler loses
track of context and may end up seeing the rest of
the program as erroneous

4
 Error Detection
Introduction

False Detection of Errors

• Take the following program:
void main () { // line 1
int i, j; // line 2
i=1; // line 3
while (i) { // line 4
int j; // line 5
…
} // line 10
j=2; // line 11
if (i<j) j++; // line 12
…
} // line 65

• If we delete the { on line 4, ‘int j’ is taken as the body of the while loop.
• This declaration of j is reported as an error (already declared in line 2)
• The close bracket on line 10 is seen as closing the main() function.
• All lines 11-65, meant to be part of ‘main()’, are reported to be “instruction located
outside a function”
• Any real errors in lines 11-65 would NOT be reported!
9

Error Detection
Introduction

Recovery from Errors

• In such cases, an intelligent compiler might
recognise that a parenthesis was missing, and look
for a place it could close the block, and then resume
processing as normal.
• Possibly indentation might give a clue.
• Or trying each location between tokens to find the
one which produces the fewest errors

10

5
 Error Detection
Introduction

Over-reporting errors
• Assume we forget to declare a variable
• If the variable is referenced 20 times, should we
receive 20 error messages? Or just 1?
• Ideally, when a reference to an undeclared variable
is found:
• The symbol table is checked to see if any
previous such error was detected for this variable.
• If yes, no message is given,
• Else:
• An error message is printed,
• The symbol table is modified to indicate this. 11

Error Detection
Introduction

Over-reporting errors
• Alternatively, the symbol table can be used to store
each occurence undeclared reference to the
variable.
• At the end of the scope, a single message is
printed:
• Undeclared reference to variable j on lines 21, 22 and 23

12

6
 Error Detection
Automatic Error Correction

Automatic Correction of Errors

• Ideally, a compiler could identify the presence of
errors, and automatically correct the code to
remove the error
• An executable could then be produced even with
source-code errors
• Time in the code-compile-debug cycle could be
saved
• BUT: if the fix provided by the compiler is wrong, it
may be difficult to locate the bug at run-time.
• For this reason, commecial compilers do not
generally offer this feature.
13

Error Detection
Automatic Error Correction

Automatic Correction of Errors

• If the compiler DOES do automatic correction, it
must document its corrections (printed warnings).
• Otherwise the programmer is debugging a program
that does not correspond to their source code!

14

7
 Error Detection
Automatic Error Correction

Some types of Error Correction: orthographic correction

• When a symbol is encountered that is not declared, the compiler can use
techniques similar to spell-checkers in word processors to guess the
correct symbol.
• Easiest to start with the symbol table, find all identifiers visible in the
current scope, and look for closest match.
• Include also list of reserved words.
• The system can try 1 character mutations:
• All chars the same except one different (more likely if keys adjacent
on keyboard)
• All chars the same except one extra
• All chars the same except for one less
• All chars the same except 2 are swapped in position
• Prefer symbols which are declared but never assigned, or assigned but
not referenced
• The parser can help: if a reserved word is expected at this point, prefer
closest reserve-word before known identifiers.
15

Error Detection
Automatic Error Correction

Some types of Error Correction:

Orthographic correction: Space insertion
• Another case is where a space char is missing.
• Here we need to test insertion of a space at each point in the
token
• The parser can help here also, informing which keywords
are expected
• For instance, with “voidmain”, and we expect amongst
other tokens “void”, then the correction is easy.

16

8
 Error Detection
Automatic Error Correction

Some types of Error Correction: Syntactic Correction

• Deletion of Token
• When a sequence of tokens does not parse, one can try
deleting each token in turn until the sequence parses.
• For instance, given
else x=0;
• …where there is no previous if statement,one could
eliminate ‘else’ and produce a correct parse.

17

Error Detection
Automatic Error Correction

Some types of Error Correction: Syntactic Correction

• Insertion of Token
• When a sequence of tokens does not parse, one can try
inserting a token in various places.
• For instance, given
if (i<j) {x=1; y=2; else x=0;
• One could try deleting “else”
• But more likely, the insertion of symbol } before the else.
• The compiler could be set such that, when an else does not
match an earlier “if”, automatically try “}” insertion.

18

9
 Error Detection
Automatic Error Correction

Error Detection: SUMMARY

• How to detect all the compilation errors in a
program.
• How to avoid false error reports.
• How to avoid reporting a cascade of spurious
errors.
• How to avoid repeating the same error message.
• How to correct errors automatically.

19

10