(This bug imported from BugSplat, Netscape's internal bugsystem. It was known there as bug #326819 http://scopus.netscape.com/bugsplat/show_bug.cgi?id=326819 Imported into Bugzilla on 02/03/99 17:02) ** Observed with 9/28 and 9/30/98 Win32 builds under US and Japanese Windows ** I was trying to reply to a long msg which I quoted in my reply and then added some lines of comments. When I engaged the English spell-checker, a crash occured. I am able to reproduce this every time I follow the steps below. 1. Select an existing msg. 2. Reply to it using HTML option. 3. Quote the original text (my auto-quote option is turned off.) 4. Write some lines (1-3 sentences should do it). 5. Engage the spell-checker. 6. See Nova crash. I have 3 separate TalkBack incident reports. I'll provide the incident numbers when they become available.

reassigning to putterman for triage.

Jeff, could you take a look at this. reassigning to ducarroz.

set qa assigned field. Kat - does it matter what the content of the quoted message is? HTML or plain text compose window?

Lisa, I can't tell without doing a lot of investigation whether or not HTML/plain text makes a difference. I wish we have Talkback reports available immediately. I think I sent at least 2 reports but I haven't found them yet. Hi, Scott, It's the msg you sent me and mscott a couple of days ago with the patch files which produces this bug every time. I described it as 'long' because it contains 2 sets of patches and I was quoting it to reply to and adding my own comments. I noticed that reaction of the editor was slowing down and that's why I figured that it might have to do with the length of the quoted msg. I tried it a few times with shorter msgs but this problem did not occur. I have a debug build with the Scott's msg in the mail folder if Jean-Francois wants to take a look. Scott should be able to send you a copy of that msg also.

Changed LAnguage to all since it occurs on both my US and Japanese Windows.

I'd appreciate someone sending me this message so that I can place the message somewhere to use to verify when this bug is fixed. Thanks.

Lisa, I just sent you a copy.

Set TFV to 4.5in

The current code of the spelling checker is limited to 32K, if you check a buffer greater than that it crash... I will change UINT2B and INT2B variable for UINT4B and INT4B to extend the max size to 0x7FFFFFFF K. Set Platform to all as the spelling checker code is cross-platform

Is this limit also imposed in the Composer? Are you changing this just for the mail compose windows? I don't want to retest spell checking from the Composer.

Yes, that would also affect the composer. This change scare me too but I am just trying to see if it's possible to do it safe.

Thanks. I appreciate it. I just wanted to make sure if your change effects Composer or not. You may wish to have Kathy Brade review your changes also since her group worked on the spell checker, I believe.

After all, the 32K limitation is not the real problem. BTW, the Mac version doesn't crash. Still looking for a fix

The crash occurs only with the HTML composer. With plain text composer it works fine.

The Mac doesn't crash because we are lucky that the memory on this platform isn't protectect. No problem to read data at the wrong location! The problem is due by the fact the HTML composer replace the quoted text with spaces during the spelling checking, in our case with more than 32K of spaces. The spelling checker doesn't support more than 32K of continuous space!

Peter, could you check if this bug occurs also on 4.06?

I'll check this out. Kat sent me the message that he used to reproduce this problem.

Kat's steps seem fairly straightforward, but I cannot even reproduce this crash on 4.5 (using 10/1a build on Windows 95). I even recreated the original message with the same content and the same attachments and not just use the forwarded email from Kat. So, I can't really try 4.06 if I can't reproduce this in 4.5 since my testing would be invalidated. I'll try this on NT 4.0 next.

Pls disregard my previous comments. It was a little harder to reproduce this bug under Windows 95. You had to type in more than three lines of text. Findings to ducarroz's original question: - Crash also occurs under 4.06 RTM build so this bug is not a regression.

I assume that this bug can be reproduced in just Composer (not mail compose). Please reassign the bug to me and my team will work on a fix for 5.0. It is *possible* that a fix Mike Judge proposed for 4.5 today might address this but I don't think it is the same area of code so I'm not counting on it.

What is the status? Is this going in by Monday? Is it a later to 5.0? Thanks, Jim the iquisitor

set tfv to 5.0; reassign to cmanske

We can't reproduce the exact circumstances without mail, but I'll test using the spell checker on very large files, once I can get it to work at all!

This bug is transferred to the wrong "product" - it's in mozilla which I thought was "mozilla classic" and no longer in use. Also, if this bug occurs under regular composer without mail, then the QA Contact should no longer be pmock@netscape.com.

Changed QA contact to sujay. As to your question on what the product should be, I'm putting Mozilla for lack of any other choice. It's not an NGLayout bug- it's browser (or maybe Ender), so until we come up with a seamonkey or browser product, what else can I do?

marking resolved won't fix in MozillaClassic codebase. If someone thinks this bug is still relevant in new seamonkey codebase, please open a new bug under Browser, or re-open this one and change the product to Browser. All open MozillaClassic bugs are being closed. Thanks.

verified in 4/27 build.