StyledEdit regression

[iam-alex]

It looks like there is recent regression in StyledEdit: hrev54330 x86_gcc2 won't open attached file, while hrev54026 x86_64 version works as intended.

Also I think that regression of StyledEdit is responsible for my broken file, fortunally it was just notes.

[humdinger]

FWIW it opens in StyledEdit with Beta2 (hrev54154+112), 32bit. Did you try a "checkfs -c" to exclude possible file system corruption?

[ttcoder]

By "broken text file", do you mean something like #10765 ? BFS disk corruption ""triggered"" by StyledEdit is very uncommon, but me and user Kev have experienced it a couple times some years ago, before beta1.

What do you mean by "won't open attached file", does the StyledEdit window open but is empty? Or it won't open at all ? Does the file open correctly with another editor, like Pe ?

If you have a reproducible test case for #10765 that would be very interesting!

Plus, what humdinger said re. checkfs, obviously!

[iam-alex]

Well, in latest hrev54352+2 I can reproduce nothing.

In first case I see incorrect atribute, prevented StyledEdit to open file. Pe opens OK. I can not reproduce that atrubute brokening.

In second case I see question marks in file while they shouldn't be. It looks like one small piece of text (truly black) had been not so corretly pasted. Pe shows questions only in that place (but, again, it was everything OK when I saved the file). I took StyledEdit as "nothing can go wrong" editor. Before.

BTW, even I've run checkfs, I do not see fs issue here.

[ttcoder]

Replying to iam-alex:

• hrev54026 x86_64 version works as intended.
• hrev54330 x86_gcc2 won't open attached file,
• in latest hrev54352+2 I can reproduce nothing.

Seems possible you are not victim of a BFS problem at all, but of a system regression that was introduced after 54026, and was fixed before 54352.

In first case I see incorrect atribute, prevented StyledEdit to open file. Pe opens OK. I can not reproduce that atrubute brokening.

In second case I see question marks in file while they shouldn't be. It looks like one small piece of text (truly black) had been not so corretly pasted. Pe shows questions only in that place (but, again, it was everything OK when I saved the file). I took StyledEdit as "nothing can go wrong" editor. Before.

BTW, even I've run checkfs, I do not see fs issue here.

If I follow correctly, there are still remaining issues with the file in new hrev. I would try to recreate the text file "from scratch" (no attributes), i.e. create a blank new file, copy-paste the old text into it, and save -- then see if the problem is gone for good in the new file.

BTW there's lots of good information about the implementation of StyledEdit and dependancies in #14674 though I can't discern a link between both tickets. Not familiar with the code though.

[iam-alex]

What I said about Haiku revisions should be carefully taken. Because initially I thought there was just fresh Unicode related regression in StyledEdit and checked the same file, namely readme.txt, on different laptops, but I didn't realize they had different atributes also.

Now, in latest Haiku, I can not reproduce when and why file attributes get corrupted, but I didn't even try it in previous hrev - because I thought problem was somewhere else.