james-mime4j-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Oleg Kalnichevski <ol...@apache.org>
Subject Re: Mime4J improvements was: Re: Thunderbird Mailbox support (patch included)
Date Fri, 03 Oct 2014 08:31:38 GMT
On Thu, 2014-10-02 at 19:37 +0200, Wolfgang Fahl wrote:
> Hi Oleg,
> Thank you for your prompt answer and advice. It worked beautifully from
> my point of view:
> https://github.com/WolfgangFahl/james-mime4j/blob/trunk/dom/src/main/java/org/apache/james/mime4j/message/BasicBodyFactory.java
> has a fix with my proposal as to make lenient Charset handling the
> default but at least switchable
> and
> https://github.com/WolfgangFahl/james-mime4j/blob/trunk/dom/src/test/java/org/apache/james/mime4j/dom/MessageCharsetLenientTest.java
> has a JUnit test that shows the modified behavior. It includes some 50
> invalid Charsets i found in my sample of 1/4 million e-mail messages.
> As far as I can tell the changes don't break any other test.

Please raise a pull request in GitHub and post a link to MIME4J-218

> The relevant bug https://issues.apache.org/jira/browse/MIME4J-218
> is still marked as resolved. Shall I add a new one or are you going to
> reopen it?

I re-opened MIME4J-218.

> With the fix above only 5 messages in my sample of 1/4 million emails
> can't be parsed by mime4j 0.8.0-SNAPSHOT. All errors are due to
> line size and header size issues. The repository above has further
> improvements on the handling of these MimeConfig settings. There seems
> to be a
> followup problem that the MimeConfig settings are not fully picked up in
> all situations. This gets visible when the exception messages has the
> current setting added e.g.
> as I did it in my changes.
> Example:
> mail<4EEB37FC.1040303@dia-bonn.de>.err:org.apache.james.mime4j.MimeIOException:
> Maximum header length limit (20000) exceeded
> E.g. when setting the maxheaderlines parameter to 20000 there are still
> situations when an exception is thrown with the maxheaderlines parameter
> being 1000. So it's seems that the config is not used consistently but
> replaced by the default in certain circumstances which I'd still have to
> debug. Is this
> worth another BR?

If you have a reasonable reproducer please raise a JIRA and attach the
test case to it.


View raw message