22315 - Entities are broken in HTML mail send

Status: VERIFIED FIXED

(MailNews Core :: MIME, defect, P3)

Description

[nhottanscp]

I used win32 build 1999-12-21-09-M13. I don't know if this is a regression from
M12.
When I sent a HTML mail with a trade mark (on windows, holding Alt and type 153
on numeric key pad). Received message contains ™ instead of ™.
As a result, the broken entity is shown in the message view as "™".

Comment 1

[nhottanscp]

On M12, mozilla freezes with the same operation. I think it is the infinite loop
fixed recently in M13.

Comment 2

[Katsuhiko Momoi]

This will have to be release noted for M12 as users trying
to input Windows-1252 only characters under Latin 1 encoding
send, Latin 1 characters under JPN send, etc. will run
into this freezing problem.

Comment 3

[rhp (gone)]

Ben,
I've noticed this with parsing some I18N messages as well.

- rhp

Comment 4

[Ben Bucksch (:BenB)]

nhotta,
If ScanHTML is fed with Unicode (neither "nbsp" nor "™"), the
problem should go away. I thought, you were working on that?

rhp,
didn't the additional else statement (mozTXTToHTMLConv.cpp, Version 1.7)
make it into M12?

Comment 5

[rhp (gone)]

> rhp,
> didn't the additional else statement (mozTXTToHTMLConv.cpp, Version 1.7)
> make it into M12?

Nope. They branched before then and I've been on my back the past 1 1/2 days
fighting the flu.

Comment 6

[nhottanscp]

>nhotta,
>If ScanHTML is fed with Unicode (neither "nbsp" nor "™"), the
>problem should go away. I thought, you were working on that?
I have proposed ScanHTML to move to nsMsgSend.cpp from nsMsgSendPart.cpp.
I do not plan to change the code myself but I can test both entity and i18n
cases once your patch is ready.

Comment 7

[Ben Bucksch (:BenB)]

Sorry, but I can't make this change, as I don't understand the data flow or the
corresponding code. I just wrote mozTXTToHTMLConv.

Comment 8

[nhottanscp]

OK, I'll try to relocate ScanHTML. In case of any internal problem in
mozTXTToHTMLConv, I'll reassign to you.

Comment 9

[nhottanscp]

Attachment: patch for the bug

Comment 10

[nhottanscp]

I attached a diff. Rich, could you review it?
I have tested HTML mails with url strings (e.g., http://home.netscape.com,
[e-mail]), an entity ™ and a simple Japanese text (copied from
bug 18410).
I don't know much about ScanHTML itself so probably more extensive testing
needed by the author ([e-mail]).

Comment 11

[nhottanscp]

The fix checked in.
For verification, please also test Japanese mail which had a problem before.

Comment 12

[Katsuhiko Momoi]

Changed QA contact to momoi as it involves intl data testing.

Comment 13

[Katsuhiko Momoi]

** Checked with 1/20/00 Win32 build **

The trademark and other HTML entities are now correctly encoded.
Also the crash I reported when I mixed Latin 1 entities with Japanese msg send
is gone from this build.