Hello to everybody
until yesterday I was an enthousiastic partecipant in the project s@h under boinc4.19, installed on over 41 hosts; I've reached 245,500 credits.
Until yesterday;
what happender?
I used notepad to open randomly one of about 600 wus thay my host have to upload and with big surprise I Found inside the wu to upload a piece of one file .doc, written with words. How it can be possible??? I disinstalled from all the pcs the boinc and sent to the trashbin all the wus to upload.
I won't continue in the partecipation.
Now I'm asking myself: during this month of number crunching, what kind of data I uploaded to the seti's servers???
Sorry for my poor english.

---

> I used notepad to open randomly one of about 600 wus thay my host have to
> upload and with big surprise I Found inside the wu to upload a piece of one
> file .doc, written with words. How it can be possible???

It is possible that BOINC picked up something that was already in RAM. Maybe it was past the end of their data structure they were grabbing, or maybe it was on disk and should not have been read.

I'm sure the developers would appreciate your help in finding the bug.

---

BOINC will swap memory in and out when you're switching between apps. Maybe Word's VM got into BOINC's swap page?

---

"I'm sure the developers would appreciate your help in finding the bug"

Are the developers reading this forum?
Anyway I'm scared to continue crunching wus and I give it up.

I won't no more do number cunching for seti@home

---

> "I'm sure the developers would appreciate your help in finding the bug"
>
> Are the developers reading this forum?
> Anyway I'm scared to continue crunching wus and I give it up.
>
> I won't no more do number cunching for seti@home

If you're done, then it doesn't matter if they do or not.

If you've thrown the work unit file away then you don't even have an example of the problem.

... so it'll stay that way until someone else sees the problem.

---

Gerone:

It would have been great if you'd checked out more results when you discovered this as well as keeping it for bug tracing. Heck, it might not even be a boinc/seti problem, but other apps you run on your system ;) A memory leak in any app could potentially produce such a side effect

---

Hi Gerone,

a thing like this could happen if your pc crashes:
The directory entry for the result file would be pointing to a record on the disk
that has not been updated yet because of the crash.
After rebooting you would see old disk data showing up in your result file.

Things like this can happen to any kind of application that writes to disk.

Did the pc having this problem crash recently?



On a side note, nobody at berkeley is reading the result files you send in.
They're analyzed automatically, and thrown away if they don't match the
other results of the WU in question.

I also don't believe that boinc can pick up data from another app in memory,
because a windows app isn't allowed to access memory that wasn't assigned
to it.


Regards Hans

---

> "I'm sure the developers would appreciate your help in finding the bug"
>
> Are the developers reading this forum?

Yes and please do post more info if you still have it.

Someone mentioned filecontent swaps. Are you running win98 or similar OS that uses FAT32? Or are you using a journalled NTFS system?
Please post information about the machine where the incident happened. Did you run any non-standard applications on it? How much RAM does it have and do you sometimes use all the RAM? Basicly any information that you would like to provide that can help the search for why this happened.

I dont know what the fuss is about. Work units have headers on them. I copied this from the start of a work unit. There is no personal information in the file. The xml tags(words)have been automatically stripped out by the browser.



20ja05ab.18441.15040.797144.66


20ja05ab
2453391.8492908
2453391.8492908
15040
0
0

20ja05ab.18441.15040.797144

12.516995221574
11.553248534959
12.592150511325
12.370102130479
1.403358366012
Fri Jan 21 08:22:59 2005
2453391.8493054

---

> I dont know what the fuss is about. Work units have headers on them. I
> copied this from the start of a work unit. There is no personal information
> in the file. The xml tags(words)have been automatically stripped out by the
> browser.
>

The 'fuss' is about the upload XML not just containing SETI XML data but apparantly containing other text that had nothing to do with SETI, namely some text from a docukment the original poster had been writing in word proccessing software.

If this has happend - and i'm sure it can in certain circumstances - then it is important to understand how and why.

---

As the admins know his email, they probably already contacted him.

Nevertheless, imho it's no fine behavior to just state such things an vanish. This boards can do no more harm then any other webpage, so even after quitting Boinc, some more info would have been possibe, even for the most paranoid persons.

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

This issue is a Tea|tempest|pot. :-)

When a file is 'deleted' only the directory entry is 'flagged/deleted'. The data that was written is still on the disk, and gets overwritten when the disk blocks are reused. If the application writes 128 bytes into a 256 byte disk block, the other 128 bytes will still have the old data in it. The application knows how much data was written and only retrieves that amount. Another aplication [ Notapad ] retrieves the whole block, as it doesn't know how much was written.

There was a flap a while ago, when a reporter noticed this and thought that M$ was collecting info from users computers.

The only way to 'correct' this is for the user to run a disk eraser program on the unused disk blocks, or the application programmers to write zeros into a disk block before they write the data.

Afterthought: This could also be a memory buffer issue. The same process as above. The application programmer could zero out a memory buffer before writing into it. I think that most programmers don't do this, as it is usually a waste of time and leads to 'code bloat'. :-)

---

> The 'fuss' is about the upload XML not just containing SETI XML data but
> apparantly containing other text that had nothing to do with SETI, namely some
> text from a docukment the original poster had been writing in word proccessing
> software.
>

Roberto didn't say that. He said "...I used notepad to open randomly one of about 600 wus".

WUs are downloaded, results are uploaded. WUs are never uploaded after being downloaded. You only have as many result files, at any one time, on your PC as you have CPUs...so he must have meant the WU files.

Between start and end of WU file there should be only text. If system crashed, many strange cross linked files can occur. But these files are never uploaded...so it wouldn't matter.

Whenever BOINC tries to talk to the servers, it first writes all its requests to a file, and then repeatedly (if necessary) tries to send this message file until sucssesful. So if one of the 1 or 2 message files were left over after a crash, I believe they are ignored. On program restart, as I recall, the code would re-create new message files.

> > The 'fuss' is about the upload XML not just containing SETI XML data but
> > apparantly containing other text that had nothing to do with SETI, namely
> some
> > text from a docukment the original poster had been writing in word
> proccessing
> > software.
> >
>
> Roberto didn't say that. He said "...I used notepad to open randomly one of
> about 600 wus".
>

Read the thread Subject: "My word file'content inside a wu to upload!!!!!"

I don't see that there is much wrong with my interpretation of the problem.

---

Yes MikeSW17 I see it your way too. Memory leak or a corrupt fielsystem has tended to be behind these types of issues before now. Shame he as gone as its a serious security issue. I have been through all my WUs on Win XP, Win server 2003 EE and Linux FC to check and I find no problems. Phew! It might be worth folk doing a regular check to see what the score is....if it was his file system then phew! ....again...may be a one-off for him and not related to Boinc anyhow. I just wonder what he would have got if he opened some random read-me file instead.....damn he should know better than just junk and run if he's responsible for running 40 odd systems. It may be we should try quite hard to find it I think?

---

Hello to all;
i do nou understand so much on all your answer;
simply talking, is it the fault of my sistem??? I opened with notepad one of the file to upload, it was about 20 Kbytes and I found that surprise.
I used a lot of Operative systems, win 98, win me, win Xp, win 2000 server sp4, win 2000 professional sp4. Intel p4 Ht, amd athlon 64, duron, normal athlon, PIII, PII, celeron and so on.
I continue to be scared to use the boinc; what could I do If I decide to partecipate again??
Sorry for my poor english and if someone answer to me, please write an easy english

Greetings
Roberto

---

""" Please post information about the machine where the incident happened. Did you run any non-standard applications on it? How much RAM does it have and do you sometimes use all the RAM? Basicly any information that you would like to provide that can help the search for why this happened.
"""

The pc is a Duron 800Mhz running win98SE; the maximun work for this pc is some daily word processing with microsoft word.
The strange is: I only checked RANDOMLY ONE WU'S RESULT and I found some word text of one of my documents inside. Considering now that I have about 245,500 credits; to reach this score I had to upload some thousands of results. Now I'm still asking myself: between these thousands of results uploaded, what kinf Od data i uploaded Too??? Is my documents archive (o f my job) in seti's server? who can assure me that I did not sent my data (and any personal data) to unknown people/organization??? At work I have a huge archive of documents and I Cannot risk my data and my documents for a program t(the boinc) that showed me to be not sure and that showed me to work as a kind of spyware. Untill now my Boinc hasn't discovered E.T. but it has at least tryied to upload my documents. Then it's more like a spyware than a research tool.

Sorry for that!

Greetings
Roberto

---

> You only have as many result files, at any one time, on your PC
> as you have CPUs...

Not true. You have as many results on your computer at one time as you have done between uploads.

---

> Untill now my Boinc hasn't discovered E.T. but it has at least tryied to
> upload my documents. Then it's more like a spyware than a research tool.

If you really believe that this is a big plot to steal word documents then you should uninstall and never run anything on your computer that communicates with the outside world.

If you think there is a chance that this is a bug, then you can help fix it.

You choose.

---

> Then it's more like a spyware than a research tool.

Sorry, but if you discovered the same problem with all results then, and only then, would I understand your concern. But you base your concern on a single result file from a single computer - even seti needs confirmation to their findings ;)

Now, what I would do in your situation: Get boinc running on a couple of systems, and for starters the one where you discovered this as well, and frequently check the result files. If the error occurs once again, report it to the developers - and if a system keeps creating that error, I would turn that one off and await a solution to the problem. And then gradualy I would add more of the systems I wanted to crunch for seti@home.

A couple of years ago I had a crash on a computer running seti classic. I discovered a wu having traces of a text file in it, so I deleted it and got a new one - I checked that one and it was a valid wu. A couple of wu's later I did a check again, and that one was valid as well.

I had another crash some months later, this time I discovered wu data in an image file (one I was working on as the crash happened). I've experinced an update for win98 causing data for one program in memory to be overwritten by Solitaire (another update to fix that was released the next day though).

Anyway, if you get my drift, to suspect boinc or seti@home to be spyware with one result as proof is not excactly a scientifically valid approach ;)

---