Some of the smilies are not displayed, i.e. the content is completely lost. The bug seems to appear espcially for smiles at column 2. I'll attach my converter test folder, so can test for yourself. Compare the msgs between 4.x and Mozilla to see the data loss.

The bug appears e.g. in msgs "smiles test 3" and "smilies" (and some others). Ignore the not recognized smilies in msgs "Try this line:" and "(no subject)".

I'll try to poke at this for M17. thanks Ben.

Mail Review hopes this is post-beta2, easy, polish work. Marking M18.

I don't think this is polish work. It seems more involved than that (?)

I also think, we shouldn't ship a product with a data loss bug, but I'm no manager :) (i.e. I don't have the overview). This seems to be some wierd logic error, a tricky off-by-one bug. Had a similar one before (is it even the same, but mscott missed the fix?).

("product" = nsbeta2)

There's a "XXX Hotfix" block in mozilla/netwerk/streamconv/convertes/mozTXTToHTMLConv, which was intended to fix this bug, when it appeared the first time. Look there first.

Nominating this for nsbeta2 as it involves data loss. This bug seem to occure in if you start a smile or wink ":)" in the 2nd column. However there may be other test cases that are not known. The best example is if you star in first column everything is fine :) this is a test :) If you start in the 2nd column :) this is a test :) The first smile is completely lost, it does not even live the litteral colon an closing parens :). However the 2nd simle still is converted. Thus we lose 2 characters. It makes me wonder if there might be other areas where loss may be occuring but we have not caught it yet.

Alan, there *are* other cases (IIRC). This is a regression caused by the performance enhancement. Adding keyword.

Thanks Ben, If you remember those other cases it would be good as the more cases of data loss we know of the more important this bug is and why it should be NSBETA2

Putting on [nsbeta2+][6/15] radar. Data loss

attaching the fix. BenB, can you give me a code review for this? Thanks.

No! lol. I didn't think it was *that* simple. Code is OK. (I'd move the Append before the STOP, but that's a minor issue that matters only for debugging.) Tested with my test cases (attached) and it works fine modulo an unrelated bug in the Mailnews flowed converter. r=me

i thought that was pretty funny too. I'll move the append before the stop. Thanks for the review.

fixed....another nsbeta2 bug bites the dust.

verified fixed win 95 moz build 2000060908. I wish the other bugs were this easy!