[jira] [Commented] (XERCESC-1984) TranscodeToStr::transcode throws an exception when transcoding to UTF-8

["Lee Doron (JIRA)" <xe...@xml.apache.org>]

    [ https://issues.apache.org/jira/browse/XERCESC-1984?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13289994#comment-13289994 ] 

Lee Doron commented on XERCESC-1984:
------------------------------------

The problem with "if(charsDone < len)" is that it increases (doubles) allocSize even if the transcoder exited for reasons other than running out of available output buffer space; that can happen if the input buffer ends with the leading character of a surrogate pair. Why increase it if there might be plenty of space?

I suggest changing the conditional to:

    if(charsDone < len && (allocSize - fBytesWritten) < 4)

This ensures that there are at least 4 bytes available, which is always enough to hold at least one more multi-byte character, so charsRead won't be 0 for lack of space. (I don't believe any encodings use more than 4 bytes for a character, right?)

Likewise, I'd change the corresponding conditional in TranscodeFromStr::transcode() from:

    if(((allocSize - fCharsWritten)*sizeof(XMLCh)) < (length - bytesDone))

to:

    if(bytesDone < length && (allocSize - fCharsWritten) < 2)

There's no reason to multiply by sizeof(XMLCh) here. However, we do need to make sure there's enough room for the largest representation we might get from a transcoder, which is 2 XMLCh entries (a surrogate pair).

These could be simplified slightly by moving the entire "if" blocks to the very beginning of each loop. At that point, we know that "charsDone < len" (or, respectively, "bytesDone < length"), and we can leave out the first clause of each conditional. It will always be skipped the first time through the loop.
                
> TranscodeToStr::transcode throws an exception when transcoding to UTF-8
> -----------------------------------------------------------------------
>
>                 Key: XERCESC-1984
>                 URL: https://issues.apache.org/jira/browse/XERCESC-1984
>             Project: Xerces-C++
>          Issue Type: Bug
>          Components: Utilities
>    Affects Versions: 3.2.0, 4.0.0
>         Environment: Bug reproducible on a Red Hat 5 based platform. The bug doesn't seem to be platform specific though.
>            Reporter: Dan PV
>              Labels: exception, transcode
>         Attachments: transtest2.cpp
>
>
> This issue relates to the bug fix for issue XERCESC-1947. There are still cases where the method will fail in providing a transcoded version without throwing an exception. See the attached "transtest2.cpp" to reproduce the issue.
> The cause seems to come from the added "if((allocSize - fBytesWritten) < (len - charsDone))" condition in "TranscodeToStr::transcode" . In my provided test case I have a string composed of 6 Japanese characters (i.e. "絞り込み検索"). Once the first call to "XMLUTF8Transcoder::transcodeTo" is done, "charsRead" will return a count of 5 XMLCh readed. Since the initial allocated buffer for this string was set to 16 bytes, the condition will check against the following values "if((16 - 15) < (6 - 5))" which avoids the reallocation of a larger buffer for the UTF-8 encoded version of the string. 
> Since the reallocation doesn't take place, the code will recall "XMLUTF8Transcoder::transcodeTo" but this time the "charsRead" count will be set to 0 because there is insufficient space in the buffer and this will trigger an exception of type "Trans_BadSrcSeq".
> I suppose that the goal of this added condition was to avoid an unnecessary reallocation of a buffer but unfortunately it doesn’t work when transcoding to variable length encoding like UTF-8. The solution is probably to simply replace the condition with "if(charsDone < len)".
> Regards,
> Dan

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

       

---------------------------------------------------------------------
To unsubscribe, e-mail: c-dev-unsubscribe@xerces.apache.org
For additional commands, e-mail: c-dev-help@xerces.apache.org