Code that has no effect or is never executed (that is, dead or unreachable code) is typically the result of a coding error and can cause unexpected behavior. Such code is usually optimized out of a program during compilation. However, to improve readability and ensure that logic errors are resolved, it should be identified, understood, and eliminated.
Statements or expressions that have no effect should be identified and removed from code. Most modern compilers, in many cases, can warn about code that has no effect or is never executed. (See MSC00-C. Compile cleanly at high warning levels.)
Noncompliant Code Example
This noncompliant code example demonstrates how dead code can be introduced into a program [Fortify 2006]. The second conditional statement, if (s)
, will never evaluate true because it requires that s
not be assigned NULL
, and the only path where s
can be assigned a non-null value ends with a return
statement.
int func(int condition) { char *s = NULL; if (condition) { s = (char *)malloc(10); if (s == NULL) { /* Handle Error */ } /* Process s */ return 0; } /* Code that doesn't touch s */ if (s) { /* This code is unreachable */ } return 0; }
Compliant Solution
Remediation of dead code requires the programmer to determine why the code is never executed and then to resolve the situation appropriately. To correct the preceding noncompliant code, the return
is removed from the body of the first conditional statement.
int func(int condition) { char *s = NULL; if (condition) { s = (char *)malloc(10); if (s == NULL) { /* Handle error */ } /* Process s */ } /* Code that doesn't touch s */ if (s) { /* This code is now reachable */ } return 0; }
Noncompliant Code Example
In this example, the strlen()
function is used to limit the number of times the function s_loop()
will iterate. The conditional statement inside the loop evaluates to true when the current character in the string is the null terminator. However, because strlen()
returns the number of characters that precede the null terminator, the conditional statement never evaluates true.
int s_loop(char *s) { size_t i; size_t len = strlen(s); for (i=0; i < len; i++) { /* Code that doesn't change s, i, or len */ if (s[i] == '\0') { /* This code is never reached */ } } return 0; }
Compliant Solution
Removing the dead code depends on the intent of the programmer. Assuming the intent is to flag and process the last character before the null terminator, the conditional is adjusted to correctly determine if the i
refers to the index of the last character before the null terminator.
int s_loop(char *s) { size_t i; size_t len = strlen(s); for (i=0; i < len; i++) { /* Code that doesn't change s, i, or len */ if (s[i+1] == '\0') { /* This code is now reached */ } } return 0; }
Noncompliant Code Example (Assignment)
In this noncompliant code example, the comparison of a
to b
has no effect:
int a; int b; /* ... */ a == b;
This code is likely a case of the programmer mistakenly using the equals operator ==
instead of the assignment operator =
.
Compliant Solution (Assignment)
The assignment of b
to a
is now properly performed:
int a; int b; /* ... */ a = b;
Noncompliant Code Example (Dereference)
In this example, a pointer increment and then a dereference occur, but the dereference has no effect:
int *p; /* ... */ *p++;
Compliant Solution (Dereference)
Correcting this example depends on the intent of the programmer. For example, if dereferencing p
was a mistake, then p
should not be dereferenced.
int *p; /* ... */ ++p;
If the intent was to increment the value referred to by p
, then parentheses can be used to ensure p
is dereferenced and then incremented. (See EXP00-C. Use parentheses for precedence of operation.)
int *p; /* ... */ (*p)++;
Another possibility is that p
is being used to reference a memory-mapped device. In this case, the variable p
should be declared as volatile
.
volatile int *p; /* ... */ (void) *(p++);
Noncompliant Code Example (if/else if)
A chain of if/else if statements is evaluated from top to bottom. At most, only one branch of the chain will be executed: the first one with a condition that evaluates to true. Consequently, duplicating a condition in a sequence of if/else if statements automatically leads to dead code.
if (param == 1) openWindow(); else if (param == 2) closeWindow(); else if (param == 1) /* Duplicated condition */ moveWindowToTheBackground();
Note that duplicating a condition violates this guideline only if the duplicate conditions always behave similarly...see a compliant solution below for a condition that is textually a duplicate but behaves differently.
Compliant Solution (if/else if)
In this compliant solution, the third conditional expression has been corrected.
if (param == 1) openWindow(); else if (param == 2) closeWindow(); else if (param == 3) moveWindowToTheBackground();
Compliant Solution (Conditional Side-Effects)
This code does not violate this recommendation, because even though the conditions are textually identical, they have different side effects, because the getc()
function advances the stream marker.
if (getc() == ':') readMoreInput(); else if (getc() == ':') readMoreInput(); else if (getc() == ':') readMoreInput();
Noncompliant Code Example (logical operators)
Using the same subexpression on either side of a logical operator is almost always a mistake. In this noncompliant code example, the rightmost subexpression of the controlling expression of each if
statement has no effect.
if (a == b && a == b) { // if the first one is true, the second one is too do_x(); } if (a == c || a == c ) { // if the first one is true, the second one is too do_w(); }
Compliant Solution (logical operators)
In this compliant solution, the rightmost subexpression of the controlling expression of each if
statement has been removed.
if (a == b) { do_x(); } if (a == c) { do_w(); }
Noncompliant Code Example (Unconditional Jump)
Unconditional jump statements typically has no effect.
#include <stdio.h> for (int i = 0; i < 10; ++i) { printf("i is %d", i); continue; // this is meaningless; the loop would continue anyway }
Compliant Solution (Unconditional Jump)
The continue statement has been removed from this compliant solution.
#include <stdio.h> for (int i = 0; i < 10; ++i) { printf("i is %d", i); }
Exceptions
MSC12-EX1: In some situations, seemingly dead code may make software resilient. An example is the default
label in a switch
statement whose controlling expression has an enumerated type and that specifies labels for all enumerations of the type. (See MSC01-C. Strive for logical completeness.) Because valid values of an enumerated type include all those of its underlying integer type, unless enumeration constants are provided for all those values, the default
label is appropriate and necessary.
typedef enum { Red, Green, Blue } Color; const char* f(Color c) { switch (c) { case Red: return "Red"; case Green: return "Green"; case Blue: return "Blue"; default: return "Unknown color"; /* Not dead code */ } } void g() { Color unknown = (Color)123; puts(f(unknown)); }
MSC12-EX2: It is permissible to temporarily remove code that may be needed later. (See MSC04-C. Use comments consistently and in a readable fashion for an illustration.)
MSC12-EX3: Unused functions and variables that are part of an exported library do not violate this guideline. Likewise, code that is never executed because it is #ifdef
ed out does not violate this guideline, on the grounds that it could be subsequently used in another application, or built on a different platform.
Risk Assessment
The presence of code that has no effect or is never executed can indicate logic errors that may result in unexpected behavior and vulnerabilities. Such code can be introduced into programs in a variety of ways and eliminating it can require significant analysis.
Recommendation | Severity | Likelihood | Remediation Cost | Priority | Level |
---|---|---|---|---|---|
MSC12-C | Low | Unlikely | Medium | P2 | L3 |
Automated Detection
Tool | Version | Checker | Description |
---|---|---|---|
Astrée | 24.04 | dead-assignment | Supported + partially checked |
CodeSonar | 8.1p0 | DIAG.UNEX.* | Code not exercised by analysis |
2017.07 | NO_EFFECT
DEADCODE
UNREACHABLE | Finds statements or expressions that do not accomplish anything or statements that perform an unintended action. Can detect the specific instance where code can never be reached because of a logical contradiction or a dead "default" in Can detect the instances where code block is unreachable because of the syntactic structure of the code | |
1.2 | CC2.MSC12 | Partially implemented | |
GCC | 3.0 | Options detect unused local variables, nonconstant static variables and unused function parameters, or unreachable code respectively. | |
Helix QAC | 2024.4 | C3110, C3112, C3307, C3404, C3426, C3427 | |
Klocwork | 2024.4 | CWARN.NOEFFECT.SELF_ASSIGN CWARN.NOEFFECT.UCMP.GE CWARN.NOEFFECT.UCMP.GE.MACRO CWARN.NOEFFECT.UCMP.LT CWARN.NOEFFECT.UCMP.LT.MACRO CWARN.NULLCHECK.FUNCNAME EFFECT MISRA.STMT.NO_EFFECT UNREACH.GEN UNREACH.RETURN UNREACH.SIZEOF UNREACH.ENUM LA_UNUSED VA_UNUSED.GEN VA_UNUSED.INIT INVARIANT_CONDITION.UNREACH | |
LDRA tool suite | 9.7.1 | 8 D, 65 D, 105 D, I J, 139 S, 140 S, 57 S | Partially implemented |
Parasoft C/C++test | 2023.1 | CERT_C-MSC12-a | There shall be no unreachable code in "else" block |
PC-lint Plus | 1.4 | 438, 474, 505, 522, 523, | Fully supported |
Polyspace Bug Finder | R2024a | Checks for:
Rec. partially covered. | |
RuleChecker | 24.04 | dead-assignment | Partially checked |
SonarQube C/C++ Plugin | 3.11 | S1764, S2589, S2583, S1116, S1172, S1763, S1862, S1065, S1656, S2754, S1751 | |
Splint | 3.1.1 | | The default mode checks for unreachable code. |
7.34 | V551, V606, V649, V779 |
Related Vulnerabilities
Search for vulnerabilities resulting from the violation of this rule on the CERT website.
CVE-2014-1266 results from a violation of this rule. There is a spurious goto fail
statement on line 631 of sslKeyExchange.c. This goto
statement gets executed unconditionally, even though it is indented as if it were part of the preceding if
statement. As a result, the call to sslRawVerify()
(which would perform the actual signature verification) becomes dead code [ImperialViolet 2014].
Related Guidelines
SEI CERT C++ Coding Standard | VOID MSC12-CPP. Detect and remove code that has no effect |
ISO/IEC TR 24772 | Unspecified Functionality [BVQ] Likely Incorrect Expressions [KOA] Dead and Deactivated Code [XYQ] |
MISRA C:2012 | Rule 2.2 (required) |
Bibliography
[Fortify 2006] | Code Quality, "Dead Code" |
[Coverity 2007] |
14 Comments
Stephen Friedl
> *p++;
This is not an example of dead code, or code with no effect, but *wrong code* - there is no chance that the developer has done this on purpose (or s/he would have done)
(and I am hard-pressed to think of an actual good reason).
So suggesting that this avoids a troublesome technique - which is by nature intentional - misses the fact that it's just wrong in every way.
One can't code secure if one doesn't actually know how C works.
Douglas A. Gwyn
If there is an actual reason, it would involve volatile qualification; I think I've seen some device driver code that did this. (Reading the data reset the interrupt, and the following code wanted to use the next sequential register address.) Certainly it is too "tricky" and should be rewritten for clarity, or else fully explained in a comment.
Chris Tapp
Is the statement "Most modern compilers can warn about code that has no effect" too wide? How many would detect the following?
Robert Seacord
I watered down the statement. It is usually not possible to be too wishy washy (in most cases). 8^)
Joe Konczal
In "Noncompliant Code Example (Dereference)", the description and the code do not match. The way the code is written "*p++;", p is dereferenced and then incremented. In order for p to be incremented and then dereferenced, as the description says, the code would need to say "*++p;".
David Svoboda
Sigh...if you know what the code does, the description does match it. But if you don't, it's easy to get the wrong impression from the description. This is a case where the code can not be described by simple English. The C binding rules parse
p+
as(p+)
, which means the increment occurs before the dereference as stated. There is a trade-off between being accurate and being short...hopefully my rewrite succeeds on both counts.Joe Konczal
Ok. I see what you mean. ANSI X3.159-1989 says that the postincrement operator is applied first to *p++, but the p++ sub-expression's value is the original value of p, so after p is incremented the previous value of p is what is dereferenced. In most cases *p++ looks as if p is dereferenced first, then incremented.
(When I use the Rich Text tab, some of the backslashes before star and plus are deleted, even after I explicitly added them in Wiki Markup, so some of the pluses are interpreted as text formatting.)
Robert Seacord
In MISRA C:2012 parlance this appears to be "dead code".
Joseph C. Sible
The statement "duplicating a condition in a sequence of if/else if statements automatically leads to dead code." isn't always true, since not every condition yields the same result when evaluated multiple times. For example, there's no dead code here:
David Svoboda
I would argue that if your conditions yield different results (as they do in your example) then your conditions are not duplicates (never mind that their text is identical), and so that statement does not apply to your code.
Aaron Ballman
I think that's what we would want it to read as, but currently it just says "duplicate" which strongly implies to me "textually identical" conditions.
David Svoboda
Technically, you're both right, the wording does allow for "textually duplicate", and could be interpreted to forbid @sible's code.
Remember that this is a recommendation not a rule, so I'm not sure that striving for the precision mandated by a rule is worthwhile. Precision is a good goal, and if it is attainable then we should go for it, so can we achieve technical precision here?
Offhand, I doubt it. Consider this variant of Joe's example:
I'll argue this example is more pragmatic, but also has the same property; which is that getc() like rand() changes internal state.
Does anyone have a good suggestion for when textually-identical conditions do and do not lead to dead code?
Aaron Ballman
Off the top of my head, I'd say it's when the condition involves a side effect or a call to a function that's not "const" (in the GCC attribute sense: https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html).
David Svoboda
I added a compliant solution based on my getc code example.