[
https://issues.apache.org/jira/browse/COMPRESS-424?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16267172#comment-16267172
]
Rohan Padhye commented on COMPRESS-424:
---------------------------------------
To be fair, the JVM does insert those bounds checks anyway in order to decide
whether or not throw an exception, so I am not sure how _considerable_ the slow
down would be with additional bounds check in Java.
That said, if several sources of {{ArrayIndexOutOfBounds}} are expected, then
indeed catching them and throwing an {{IOException}} instead would make the
implementation consistent with the API specification. My primary concern is
that applications that process user-derived input files should expect the
decompression to either succeed or to fail using an {{IOException}}, such that
they can handle either case in an application-specific way (e.g. show an error
message to the end-user). Throwing an undocumented unchecked exception would
probably lead such an application to crash unexpectedly or log the error in a
top-level exception handling mechanism instead of gracefully interacting with
the end-user. Consequently, developers using the bzip2 API would be forced to
explicitly handle {{ArrayIndexOutOfBounds}}, but this seems like an
implementation detail that is being leaked outside.
The risk with this approach, of course, is that any {{ArrayIndexOutOfBounds}}
would be wrapped in an {{IOException}}, even if there was a genuine bug in
processing a well-formed file (and would thus make such a hypothetical bug
slightly harder to detect).
> [bzip2] Multiple ArrayIndexOutOfBoundsException(s) when decompressing
> malformed input
> -------------------------------------------------------------------------------------
>
> Key: COMPRESS-424
> URL: https://issues.apache.org/jira/browse/COMPRESS-424
> Project: Commons Compress
> Issue Type: Bug
> Components: Compressors
> Affects Versions: 1.14, 1.15
> Reporter: Rohan Padhye
> Priority: Minor
> Attachments: bad1.bz2, bad2.bz2, bad3.bz2
>
>
> Encountered multiple unchecked exceptions thrown from
> {{BZip2CompressorInputStream.<init>}} when parsing malformed files.
> {{ArrayIndexOutOfBoundsException}} is an unchecked exception that is not
> documented in this API; therefore, such exceptions can cause stability issues
> in applications that are not expecting them. Instead, an {{IOException}}
> should be thrown indicating that the input stream contains malformed data.
> Stack traces for three distinct (but possibly related) sources of exceptions
> follow:
> {noformat}
> java.lang.ArrayIndexOutOfBoundsException: 65536
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.hbCreateDecodeTables(BZip2CompressorInputStream.java:422)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.createHuffmanDecodingTables(BZip2CompressorInputStream.java:546)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.recvDecodingTables(BZip2CompressorInputStream.java:518)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.getAndMoveToFrontDecode(BZip2CompressorInputStream.java:555)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.initBlock(BZip2CompressorInputStream.java:324)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.<init>(BZip2CompressorInputStream.java:135)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.<init>(BZip2CompressorInputStream.java:112)
> {noformat}
> {noformat}
> java.lang.ArrayIndexOutOfBoundsException: 6
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.recvDecodingTables(BZip2CompressorInputStream.java:493)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.getAndMoveToFrontDecode(BZip2CompressorInputStream.java:555)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.initBlock(BZip2CompressorInputStream.java:324)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.<init>(BZip2CompressorInputStream.java:135)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.<init>(BZip2CompressorInputStream.java:112)
> {noformat}
> {noformat}
> java.lang.ArrayIndexOutOfBoundsException: 18002
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.getAndMoveToFrontDecode(BZip2CompressorInputStream.java:605)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.initBlock(BZip2CompressorInputStream.java:324)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.<init>(BZip2CompressorInputStream.java:135)
> at
> org.apache.commons.compress.compressors.bzip2.BZip2CompressorInputStream.<init>(BZip2CompressorInputStream.java:112)
> {noformat}
> The inputs were found by mutating random bytes in a simple well-formed file
> (a compressed string of zeros).
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
