Fix crash during Search result

[donho]

Fix #11883 partially.

[donho]

@Yaron10

The second search-results are not lexed properly, and NPP may even crash.

I have reproduced the crash, and fixed it (hopefully).
Could you check the binary to validate this point (only crash) please?

[donho]

@Yaron10
As soon as the confirmation of fix of crash issue is done,
I'll push the fix of Search Result's syntax highlighting issue.

[Yaron10]

@donho,

The crash was not consistent (about 1 of 10 NPP startups).
Unfortunately I've just encountered a crash with the new binary.

Thank you for looking into it.

[Yaron10]

Unfortunately I've just encountered a crash with the new binary.

Positive.

I might be wrong, but the crash seems to happen more frequently in NPP x32 (in x64 system).

[Yaron10]

@ArkadiuszMichalski,

Could you please try reproducing the crash on your machine?

[ArkadiuszMichalski]

Crash still exist:

Notepad++ v8.4.3   (32-bit)
Build time : Jul  8 2022 - 15:52:51
Path : E:\Download\_chrome\npp.8.4.3.portable\notepad++.exe
Command Line : -multiInst
Admin mode : OFF
Local Conf mode : ON
Cloud Config : OFF
OS Name : Windows 11 (64-bit) 
OS Version : 21H2
OS Build : 22000.778
Current ANSI codepage : 1250
Plugins : 
    mimeTools (2.8)
    NppConverter (4.4)
    NppExport (0.4)

[Yaron10]

@ArkadiuszMichalski,

Thank you.

[donho]

@Yaron10 @ArkadiuszMichalski
I cannot reproduce the crash at all with v8.4.3 x32 official portable version (neither with x32 Debug mode)

Could you provide a way to reproduce crash steadily please?

[ArkadiuszMichalski]

@donho

1. Open Notepad++
2. Open some file, like license.txt
3. Go to Find in Files (Ctrl+Shift+F) and mark Follow current doc.
4. Find some text.

If crash not happend after two searches close and run NPP again and repeat above steps. Currently I get crash after first run.

[xomx]

@ArkadiuszMichalski
I can reproduce the crash with your steps, but on the 32-bit N++ only, x64 does not crash.

Edited:
And now I cannot reproduce it even on the very same binaries (I got the crash before on the v8.4.3 x32 official portable version and also on my 32-bit Debug version of the current GitHub master) , weird!
Thankfully, I did a screenshot of the crash before, so this is everything I have for now:

[ArkadiuszMichalski]

@xomx Try with version from this PR (looks like it breaks more often than the official one). If after catching a crash and after a few attempts, there is still no crash, try run system again (in the past I had a case that after a few crashes in NPP they stopped appearing until the system restarted).

[Yaron10]

@donho,

Could you provide a way to reproduce crash steadily please?

That's the problem: it's inconsistent and can not be reproduced steadily.

You're trying to separate the crash from the wrong lexing. - Your knowledge and understanding are better than mine.
Still, allow me to ask: did you find out why in "Find in Files" the lexing breaks after the first search? Perhaps it IS related to the random illusive crash?

@ArkadiuszMichalski has mentioned an important point:

If crash not happend after two searches close and run NPP again and repeat above steps.

And also, try it with the x32 build.

Thank you.

[donho]

Try with version from this PR (looks like it breaks more often than the official one).

It should be more stable though. In anyway, I believe it doesn't contribute to the crash.

• Open Notepad++

• Open some file, like license.txt

• Go to Find in Files (Ctrl+Shift+F) and mark Follow current doc.

• Find some text.

Yes, now I can reproduce the crash with x86 official release. However, in debug mode (the same source code) I can see the syntax highlighting is lost, but no crash at all.

It seems it's rather the race condition on Scintilla - the execution on debug binary is much slower; since Scintilla is not thread safe, when the execution speeds up, there might be several accesses to Scintilla, which lead crash.

Since I cannot find the crash reason from the debug, I'll revert 0b5785c to see if the crash still happens.

[Yaron10]

I'll revert 0b5785c to see if the crash still happens.

We can all try it now with the previous commit.

Would you mind waiting a couple of days?
You or @xomx might still come up with a solution. :)

[donho]

Found it!
I'll do a commit in 30 minutes!

[donho]

The crash should be fixed in the latest commit.
There is still the colorizing problem though.
Once the crash-fix is confirmed, I will revert 0b5785c to fix the rest of problem.

[Yaron10]

@donho,

The crash seems to be fixed. 👍
Thank you.

I suppose you've noticed that with the recent commit the results in "Find in Files" are lexed but wrongly.

I will revert 0b5785c to fix the rest of problem.

Do you mean back to "no folding"?

[Yaron10]

And the lexing in "Find All in Current Document" is broken.

[donho]

@Yaron10
Sorry my bad. I was too excited to commit the correct fix.
I just committed the correct one. Please check the latest commit. Sorry again for the error.

[ArkadiuszMichalski]

Crash was fixed.

[Yaron10]

@donho,

Perfect! 👍
Thank you for you time and work.

I'm sharing your excitement. :)

So there's no need to revert 0b5785c, correct?

[donho]

@ArkadiuszMichalski @Yaron10
Thank you for the confirmation!

So there's no need to revert 0b5785c, correct?

Indeed, it's not necessary to revert this PR.
OTOH, I remark there's still the folding performance issue (on the 2nd search) when "color" is searched in all file from root of Notepad++ project.
I may try to improve the method ScintillaEditView::collapse() in another PR.

[Yaron10]

Indeed, it's not necessary to revert this PR.

That's great.

OTOH, I remark there's still the folding performance issue (on the 2nd search) when "color" is searched in all file from root of Notepad++ project.

There's a performance issue in the first search too. :)
#11774.

I may try to improve the method ScintillaEditView::collapse() in another PR.

👍
Thank you and best of luck.