Build 3245: need your help to find bugs

Hello, All,

In build 3245 just released we have significantly reworked the internals of VFS implementation (the reason is the huge delays when working with Idea on linux with lots of automounted disks). For the whole of last week I've been struggling with bugs caused by this refactoring. While I did my best to get it bug free, there is a horribly large possibility of some bugs being still there. I would greatly appreciate your help in pointing this out.

Thank you in advance,
Eugene.

21 comments
Comment actions Permalink

The first bug: 5245, not 3245 ;)

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

It only shows there are others as well...:)

0
Comment actions Permalink

How do I test the VFS implementation? What kind of operations would be affected by the VFS implementation? What kind of symptoms could bugs there have? Wrong file locations? Incorrect file contents? Data loss? Exceptions? Something else? How would I know this is caused by the VFS implementation?
I have basically no idea how to help you test this. Even only a few examples of bugs you have already fixed would help greatly.

Bas

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

All you mentioned and more. The bugs I have already encountered include duplicate classes in the project view, different assertions from write psi actions and slow file synchronization.

0
Comment actions Permalink

This might be related. I get the exceptions below every time I open my project.

Bas

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

Nope, this is not related, I'll look at this on Monday (since this exception will be assigned to me anyway:)

Eugene.

0
Comment actions Permalink

It would be interesting to get guess of the possibility of
data-corruption or loss, so we can plan on "backups" - otherwise I'd
tend to not use 5245 "really" (?) but then nothing will be tested... I
found this thread rather "frightening" in that I really thought about
not upgrading to 5245 - curiosity won, however ;)
Still would appreciate the guess.

regards,

Messi

Eugene Vigdorchik wrote:

All you mentioned and more. The bugs I have already encountered include duplicate classes in the project view, different assertions from write psi actions and slow file synchronization.

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

Just a guess, but... Is your project referencing idea.jar? If this is the case, then the reason of the problem is probably our obfuscator.

0
Comment actions Permalink

Yes, you guessed right, it's a plugin.

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

This is fixed.

0
Comment actions Permalink

Cool, thanks Eugene!

0
Comment actions Permalink

Eugene Vigdorchik wrote:

Hello, All,

In build 3245 just released we have significantly reworked the
internals of VFS implementation (the reason is the huge delays when
working with Idea on linux with lots of automounted disks). For the
whole of last week I've been struggling with bugs caused by this
refactoring. While I did my best to get it bug free, there is a
horribly large possibility of some bugs being still there. I would
greatly appreciate your help in pointing this out.

Thank you in advance, Eugene.


Hello eugene, this sounds exciting, do you plan to use FAM and inotify
as well? I've written a script that simulates OSX fslogger for
inotify-enabled kernels.

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

There are no plans to support native watching with linux in 6.0.

0
Comment actions Permalink

What if I duplicate the fslogger output exactly so you don't need to
write any new code?

Eugene Vigdorchik wrote:

There are no plans to support native watching with linux in 6.0.

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

That would definitely make my life easier:)
But how your script is going to be distributed/setup?

0
Comment actions Permalink

Eugene Vigdorchik wrote:

That would definitely make my life easier:)
But how your script is going to be distributed/setup?


We would kill for this functionality here so we would be willing to do
all the work for setup, support, etc. We would be okay if this were
secret feature disabled by default. The script depends on inotify and
python-inotify, which are generally not available and/or installed by
default for all linuxes.

I could make fslogger-compatible, but currently I provide simpler interface:
C xx.xxx = file created
M xx.xxx = file modified
D xx.xxx = file deleted

fslogger looks more complex. However if you want compatibility I will do it.

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

Can your script check for all prerequisites like inotify itself? If yes then it would be safe to distribute it with idea and only watch with its help when it explicitly says so.

0
Comment actions Permalink

What's a prerequisite?

Eugene Vigdorchik wrote:

Can your script check for all prerequisites like inotify itself? If
yes then it would be safe to distribute it with idea and only watch
with its help when it explicitly says so.

0
Avatar
Eugene Vigdorchik
Comment actions Permalink

Oops, sorry for my bad English.
I meant could the script check for kernel inotify-enableness and possibly other stuff it needs itself?

0
Comment actions Permalink

Eugene Vigdorchik wrote:

Oops, sorry for my bad English.
I meant could the script check for kernel inotify-enableness and possibly other stuff it needs itself?


Yes, the script would simply quit or print some error code if it doesn't
work.

0

Please sign in to leave a comment.