Page History

The abstract InputStream.read() and Reader.read() methods are used to read a byte or character, respectively, from a stream. The  InputStream.read() method reads a single byte from an input source and returns its value as an int in the range 0 to 255 (0x00-0xff). The Reader.read() method reads a single character and returns its value as an int in the range 0 to 65,535 (0x0000-0xffff). Both methods return the 32-bit value -1 (0xffffffff) to indicate that the end of the stream has been reached and no data is available. The larger int size is used by both methods to differentiate between the end-of-stream indicator and the maximum byte (0xff) or character (0xffff) value. The end-of-stream indicator is an example of an in-band error indicator. In-band error indicators are problematic to work with, and the creation of new in-band-error indicators is discouraged.

Prematurely converting the resulting int to a byte or char before testing for the value −1 makes it impossible to distinguish between characters read and the end of stream indicator. Programs must check for the end of stream before narrowing the return value to a byte or char.

This rule applies to any method that returns the value −1 to indicate the end of a stream. It includes any InputStream or Reader subclass that provides an implementation of the read() method. This rule is a specific instance of NUM12-J. Ensure conversions of numeric types to narrower types do not result in lost or misinterpreted char type is the only unsigned primitive type in Java. It is easy to overlook this fact and assume that a signed value can be stored and retrieved successfully. Common effects of the defective code include memory leaks and misrepresented data.

Noncompliant Code Example

...

This noncompliant example is from the {{sun.net.httpserver.ChunkedInputStream}} class. The {{InputStream}} class's {{read()}} method returns a signed byte in the form of a signed integer. In this case, the end of stream is being checked by casting the return value to a {{char}}. This conversion would leave the value of {{c}} as {{0xffff}} ({{Character.MAX_VALUE}}, decimal 65535) instead of -1. The termination test is doomed to fail. \[[Pugh 08|AA. Java References#Pugh 08]\] 

(byte)

FileInputStream is a subclass of InputStream. It will return −1 only when the end of the input stream has been reached. This noncompliant code example casts the value returned by the read() method directly to a value of type byte and then compares this value with −1 in an attempt to detect the end of the stream.

FileInputStream in;
// Initialize stream 
byte data;
while ((data = (byte) in.read()) != -1) { 
  // ... 
}

If the read() method encounters a 0xFF byte in the file, this value becomes indistinguishable from the −1 value used to indicate the end of stream, because the byte value is promoted and sign-extended to an int before being compared with −1. Consequently, the loop will halt prematurely if a 0xFF byte is read.

Compliant Solution (byte)

Use a variable of type int to capture the return value of the byte input method. When the value returned by read() is not −1, it can be safely cast to type byte. When read() returns 0x000000FF, the comparison will test against 0xFFFFFFFF, which evaluates to false.

FileInputStream in;
// Initialize stream 
int inbuff;
byte data;
while ((inbuff = in.read()) != -1) { 
  data = (byte) inbuff;
  // ...  
}

Noncompliant Code Example (char)

FileReader is a subclass of InputStreamReader, which is in turn a subclass of Reader. It also returns −1 only when the end of the stream is reached. This noncompliant code example casts the value of type int returned by the read() method directly to a value of type char, which is then compared with −1 in an attempt to detect the end of stream. This conversion leaves the value of data as 0xFFFF (e.g., Character.MAX_VALUE) instead of −1. Consequently, the test for the end of file never evaluates to true.

FileReader in;
// Initialize stream 
char data;
while ((data = (char) in.read()) 

char c;
while ((c=(char)in.read())!= -1) { 
  // ... 
}

Compliant Solution (char)

Always use Use a signed type of sufficient size to store signed data. To be compliant, use an integer type to check for EOF while reading in datavariable of type int to capture the return value of the character input method. When the value returned by read() is not −1, it can be safely cast to type char.

FileReader in;
// Initialize stream 
int inbuff;
char cdata;
while ((cinbuff = in.read()) != -1) { 
  data = (char) inbuff;
  // ...  
}

Risk Assessment

Trying to store signed data in an unsigned type can lead to misinterpretations about the actual value.Historically, using a narrow type to capture the return value of a byte input method has resulted in significant vulnerabilities, including command injection attacks; see CA-1996-22 advisory. Consequently, the severity of this error is high.

Automated Detection

...

TODO

Related Vulnerabilities

Search for vulnerabilities resulting from the violation of this rule on the CERT website.

References

Some static analysis tools can detect violations of this rule.

Tool	Version	Checker	Description
Parasoft Jtest	Include Page Parasoft_V Parasoft_V	CERT.FIO08.CRRV	Check the return value of methods which read or skip input
SpotBugs	Include Page SpotBugs_V SpotBugs_V	EOS_BAD_END_OF_STREAM_CHECK	Implemented (since 4.4.0)

Related Guidelines

Bibliography

...

Image Added Image Added Image Added Wiki Markup\[[API 06|AA. Java References#API 06]\] Class {{InputStream}} \[[JLS 05|AA. Java References#JLS 05]\] 4.2 Primitive Types and Values \[[Pugh 08|AA. Java References#Pugh 08]\] "Waiting for the end"