quick question for someone in-the-know.

What stops a project from going into an infinite loop, never completing, and putting boinc into stasis?

I noticed that einstein wu's, once started, seem to keep increasing the 'to completion' field in the boinc manager for quite some time, before trending down to completion. By extrapolation, a deviant could keep increasing that field and sit in running mode 'forever'. I don't think that the deadline date stops the process either, because my slow machines sometimes miss the deadline for various reasons and never get terminated as a result.

The abort button! :)

quick question for someone in-the-know.

What stops a project from going into an infinite loop, never completing, and putting boinc into stasis?
...

Each WU has <rsc_fpops_bound>, which tells BOINC the maximum time it can be allowed to process. For MB work here, it's 10 times the raw estimate, for AP IIRC it's 25 times. If the task isn't completed within that limit, BOINC aborts it with a "Maximum CPU time exceeded" message.

That doesn't totally eliminate the possibility that a work creation mistake might produce WUs that loop forever and have a huge rsc_fpops_bound, but it's highly unlikely. A malicious project might do it deliberately, though, so users should beware that possibility.

As to Einstein showing increasing time at the beginning, that's probably a simple matter of the estimated progress it provides to BOINC not being linear. It's difficult to get that right.

                                                             Joe

wow, thanks.

Seems like a hole in boinc's design, as long the possibility exists to loop forever. Have more elegant schemes been suggested to limit this possibility? I have a couple of half-baked ideas but won't bog the thread down here if someone has actually thought more deeply about this.

wow, thanks.

Seems like a hole in boinc's design, as long the possibility exists to loop forever. Have more elegant schemes been suggested to limit this possibility? I have a couple of half-baked ideas but won't bog the thread down here if someone has actually thought more deeply about this.

How do you tell?

The applications for CPDN run for months.

There are applications that only update the % complete once from 0% to 100%.

Now combine the two (not that this has happened yet).

---

BOINC WIKI

A malicious project might do it deliberately, though, so users should beware that possibility.

A truly malicious project could send out work units that will never validate or always generate client errors. They don't have to infinite loop the computer.

It would surprise if this hasn't already happened in error on a project where the science part wasn't properly debugged.

---

I thought you were talking about Apple HQ :-)

1 Infinite Loop
Cupertino, CA

---

A malicious project might do it deliberately, though, so users should beware that possibility.

A truly malicious project could send out work units that will never validate or always generate client errors. They don't have to infinite loop the computer.

It would surprise if this hasn't already happened in error on a project where the science part wasn't properly debugged.

It has happened to most of the projects at least once. It is sort of why ATA tends to go in first...

---

BOINC WIKI

A malicious project might do it deliberately, though, so users should beware that possibility.

A truly malicious project could send out work units that will never validate or always generate client errors.

A truly malicious project is going grant credit based on the amount of spam their "science application" can send out.

---