8890 - High ascii in the attachement are garbled

Status: VERIFIED FIXED

(MailNews Core :: Internationalization, defect, P3)

Description

[marina]

Steps to reproduce:
-sent to your acoount several messages from 4.x with attachements on different
lang (french, spanish, german,danish);
-view messages in 5.0;
//note: all high ascii in the attachements are garbage
(observed with 6/25/99 build)

Comment 1

[marina]

I have to add some comments to this bug report:  high ascii are garbled in the
body when it was sent In-line ("Send page"). In case i forward the same message
with incorrectly displayed chars to myself they are rendering correctly.

Comment 2

[nhottanscp]

Using 4.5, I send page home.netscape.com/fr to myself. I can view the page with
5.0 (local build updated/build this morning) correctly.

Comment 3

[marina]

I looked at the source of pages that have a problem and can add some info : all
problematic pages have in the source this line:"Disposition :inline".Those that
display fine do not mention anything about inline disposition.

Comment 4

[Katsuhiko Momoi]

After looking this over with Marina, the following seems to be
causing this problem.

When the original HTML page was sent as attachment from 4.5, the
MIME option was set to "8-bit as is" and it was HTML mail.

The above URL contains both HTML entities and 8-bit characters,
and it arrives with the following headers:

Charset: iso-8859-1
Content-Transefr encoding: 8-bit

It's this type of msgs we seem to have a problem displaying. I was able
to reproduce the problem under this condition.

As an experiment, I sent 8-bit Latin 1 characters in plain text
mail body and the display was OK.
It then seems that we have a problem with 8-bit characters in HTML
body.

Comment 5

[nhottanscp]

Accepting, I can reproduce this. Looks like the conversion is done from
ISO-8859-1 to UTF-8 in libmime. Need more investigation.
Adding [e-mail] to cc.

Comment 6

[nhottanscp]

My last comment is incorrect. The conversion I saw was done for the subject
header not for the attachment.
Charset label of multi part is not used for the conversion (main body charset is
used instead) in libmime.
Reassign to Rich.

--------------E29790070C6792C87CB96470
Content-Type: text/html; charset=iso-8859-1;
 name="article.jhtml"
Content-Transfer-Encoding: 8bit
Content-Disposition: inline;
 filename="article.jhtml"
Content-Base: "http://cnn.passagen.se/ekonomi/article
        .jhtml?articleID=556254"
Content-Location: "http://cnn.passagen.se/ekonomi/article
        .jhtml?articleID=556254"

Comment 7

[Katsuhiko Momoi]

Thanks, Naoki. When I included 8bit characters in the 1st body/8-bit
plain text main and then attached the above URL page, then both
the 1st body part and the attachment body showed 8-bit characters
correctly.

Comment 8

[rhp (gone)]

Looking into it...

Comment 9

[rhp (gone)]

Good catch and thanks for the help! I checked in a fix this afternoon
(Saturday).

- rhp

Comment 10

[Katsuhiko Momoi]

To marina for verification.

Comment 11

[Katsuhiko Momoi]

** Checked witrh 7/12/99 Win32 build **

(marina is on vacation -- pinch-hitting ...)

I sent the original problem page with 4.7 with no 8-bit characters
in the main body. This created a multi-part message with the
1st part bearing charset=us-ascii and the 2nd part, charset=iso-8859-1.

The above had no problem displaying the accented characters
of the attachment.

Marking the fix verified.