http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=18550193

I noticed the same error on a WU on Main last week. It was also on an Android app, although over there it's still 7.21.

---

David
signature sent back to alpha testing

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=18550193

I noticed the same error on a WU on Main last week. It was also on an Android app, although over there it's still 7.21.

is this a known error?

Yes:

http://setiweb.ssl.berkeley.edu/beta/forum_thread.php?id=2164&postid=53151

and more recently:

[boinc_projects] If your project has Android apps, please read!
David Anderson davea at ssl.berkeley.edu
Tue Jan 13 14:00:13 PST 2015

Previous message: [boinc_projects] Job limits not respected
Next message: [boinc_projects] debugging graphics apps
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

--------------------------------------------------------------------------------

Second notice:
Android 5.0 (which is currently shipping, and will soon be the majority version)
supports ONLY position-independent executable (PIE) native-mode apps.

This means that in order to work with Android 5 devices,
you'll need to rebuild your apps with appropriate options;
instructions for doing this are here:
http://boinc.berkeley.edu/trac/wiki/AndroidBuildApp#Position-independentexecutablesPIE

Even before you do this, you should associate your current Android apps
with a plan class that prevents them from being sent to Android 5.0+ devices;
otherwise you'll get a lot of failed jobs
and there will be a lot of frustrated volunteers.

The following projects are affected:

Asteroids at home
Collatz Conjecture
Einstein at home
Enigma at home
GPUGrid.net
Moo! Wrapper
OProject at Home
Quake Catcher Network
Rosetta at home
SETI at home
theSkyNet POGS
World Community Grid
Yoyo at home

If you have any questions, or need help rebuilding your apps,
please let me know.

-- David

Claggy

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=18550193

I noticed the same error on a WU on Main last week. It was also on an Android app, although over there it's still 7.21.

is this a known error?

Yes:

http://setiweb.ssl.berkeley.edu/beta/forum_thread.php?id=2164&postid=53151

and more recently:

[boinc_projects] If your project has Android apps, please read!
David Anderson davea at ssl.berkeley.edu
Tue Jan 13 14:00:13 PST 2015

--------------------------------------------------------------------------------

Second notice:
Android 5.0 (which is currently shipping, and will soon be the majority version)
supports ONLY position-independent executable (PIE) native-mode apps.

This means that in order to work with Android 5 devices,
you'll need to rebuild your apps with appropriate options;

Even before you do this, you should associate your current Android apps
with a plan class that prevents them from being sent to Android 5.0+ devices;
otherwise you'll get a lot of failed jobs
and there will be a lot of frustrated volunteers.

The following projects are affected:

SETI at home

If you have any questions, or need help rebuilding your apps,
please let me know.

-- David

Claggy

So apparently Seti, the project from which Boinc was born, has NOT done so yet?

---

David
signature sent back to alpha testing

So then what happened to this one? http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=6888453

Not all of the computers that worked on it were even Androids. And it's not the missing autocorr problem either.

---

David
signature sent back to alpha testing

So then what happened to this one? http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=6888453

Not all of the computers that worked on it were even Androids. And it's not the missing autocorr problem either.

Well, http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=70035 Owner: Eric J Korpela who was marked invalid might be able to shed some light on it!

---

So then what happened to this one? http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=6888453

Not all of the computers that worked on it were even Androids. And it's not the missing autocorr problem either.

If you download the Wu, the SETI fanout url generator is one of the files hosted at Lunatics, use it to generate the url, then just download the Wu,

http://setiathome.berkeley.edu/beta/download/370/27my12ac.2218.11118.140733193388043.16.76.vlar

and check the header, it says:

<autocorr_thresh>1000</autocorr_thresh>
<autocorr_per_spectrum>0</autocorr_per_spectrum>
<autocorr_fftlen>8</autocorr_fftlen>

Same as the 'WUs with very unusual Autocorr parameters' Joe posted about.

Claggy

So then what happened to this one? http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=6888453

Not all of the computers that worked on it were even Androids. And it's not the missing autocorr problem either.

It does have the bad autocorr parameters:

<autocorr_thresh>1000</autocorr_thresh>
<autocorr_per_spectrum>0</autocorr_per_spectrum>
<autocorr_fftlen>8</autocorr_fftlen>

The 5 "Completed, marked as invalid" and the 1 "Completed, can't validate" all appear to have crunched the WU as directed. The parameters mean there cannot be a best_autocorr, and the Validator rejects the results because there isn't one.

The 2 opencl_ati5 tasks which errored did so on a kernel related to autocorr processing. The armv7-vfpv4 error 151 may not be related as it did not happen until a restart at nearly 75% progress.

                                                                  Joe

Okay, thanks, guys. Based on your explanations, I will probably add 27my12 to my list of tasks I abort if I see them.

[edit]
But having said that, I'm now looking at others from that date and seeing some that finished and validated. There's one where I got an invalid and some other hosts got invalids or errors, but two hosts did end up validating. So I guess I won't dump them after all.

---

David
signature sent back to alpha testing