Re: [boinc_dev] BOINC API issue with android arm?

[David E Kim]

I rebuilt everything using android_api_dpa and then tested it on our devices 
and it looks like our app is getting stuck on suspend and there are seg fault 
issues when trying abort, quit, suspend.  Weird stuff happening.

I’ll take a closer look at the code changes and try to figure out what is 
happening when I get a chance.

David K




> On Apr 10, 2017, at 2:16 AM, Oliver Bock  wrote:
> 
> Hi David A.
> 
> On 08.04.17 2:13 , David Anderson wrote:
>> I made some revisions to David Kim's code and put it in a branch:
>> https://github.com/BOINC/boinc/tree/android_api_dpa
> 
> Can we please use the PR David K. thankfully created for this purpose?
> 
> https://github.com/BOINC/boinc/pull/1858
> 
>> The main changes:
> 
> PRs are meant to facilitate discussion *and* documentation like the one
> in this mail thread but next to the source code and the suggested patch
> itself. It gives the discussion much better context and retains code
> authorship.
> 
> Should you (D.A.) decide against this then please add all of these
> emails and references to your branch and/or commits to the related issue:
> 
> https://github.com/BOINC/boinc/issues/1850
> 
> For the future it would be highly appreciated if our workflow support
> tool of choice, GitHub, would be used for all of this. Leaving the
> toolchain makes development and code archeology harder and requires more
> effort than necessary - even your own.
> 
> Best,
> Oliver
> 

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Oliver Bock]

Hi David A.

On 08.04.17 2:13 , David Anderson wrote:
> I made some revisions to David Kim's code and put it in a branch:
> https://github.com/BOINC/boinc/tree/android_api_dpa

Can we please use the PR David K. thankfully created for this purpose?

https://github.com/BOINC/boinc/pull/1858

> The main changes:

PRs are meant to facilitate discussion *and* documentation like the one
in this mail thread but next to the source code and the suggested patch
itself. It gives the discussion much better context and retains code
authorship.

Should you (D.A.) decide against this then please add all of these
emails and references to your branch and/or commits to the related issue:

https://github.com/BOINC/boinc/issues/1850

For the future it would be highly appreciated if our workflow support
tool of choice, GitHub, would be used for all of this. Leaving the
toolchain makes development and code archeology harder and requires more
effort than necessary - even your own.

Best,
Oliver

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David Anderson]

I added it.
Uncomment the #define MSGS_FROM_FILE,
then put msgs like "" in msgs.txt.
When the API sees this, it will handle it and delete the file.

-- David

On 4/7/2017 5:28 PM, David E Kim wrote:

I’ll can test this.  That’s the only way I’d know if it’s ok.

It will be easier for me to test if you include the text file messaging if you 
don’t mind/ or I’d just re-implement the signals.

Thanks David!

David K





On Apr 7, 2017, at 5:13 PM, David Anderson  wrote:

I made some revisions to David Kim's code and put it in a branch:
https://github.com/BOINC/boinc/tree/android_api_dpa

The main changes:

- I added and fixed comments about the thread/signal structure

- In exit_from_timer_thread():
  removed the "boinc_status.suspended = false" and the first "sleep(2.0)".
  I don't see why either of these is needed.

- in timer_handler():
  If detect heartbeat failure, don't bother sending signal to work thread;
  exit_from_timer_thread() already does this.

- in worker_signal_handler, don't bother checking that we're the timer thread.
  Now that we use pthread_kill() instead of setitimer(),this will always be the 
case.

- I didn't include the stuff about debugging using external signals.
  That's a good idea, but there's a simpler way to do it:
  have the API look for control messages in a file instead of shared memory.
  e.g. you could tell the app to quit by writing  to the file.
  This would work on Win too.
  (I didn't add this, but will on request; it's a two-line change)

- didn't include the "in_critical_section > 0" changes; not needed.

David K, can you take a look at this and see if it's OK?

Thanks -- David


___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David Anderson]

I'll add the text-file way now.
-- D

On 4/7/2017 5:28 PM, David E Kim wrote:

I’ll can test this.  That’s the only way I’d know if it’s ok.

It will be easier for me to test if you include the text file messaging if you 
don’t mind/ or I’d just re-implement the signals.

Thanks David!

David K





On Apr 7, 2017, at 5:13 PM, David Anderson  wrote:

I made some revisions to David Kim's code and put it in a branch:
https://github.com/BOINC/boinc/tree/android_api_dpa

The main changes:

- I added and fixed comments about the thread/signal structure

- In exit_from_timer_thread():
  removed the "boinc_status.suspended = false" and the first "sleep(2.0)".
  I don't see why either of these is needed.

- in timer_handler():
  If detect heartbeat failure, don't bother sending signal to work thread;
  exit_from_timer_thread() already does this.

- in worker_signal_handler, don't bother checking that we're the timer thread.
  Now that we use pthread_kill() instead of setitimer(),this will always be the 
case.

- I didn't include the stuff about debugging using external signals.
  That's a good idea, but there's a simpler way to do it:
  have the API look for control messages in a file instead of shared memory.
  e.g. you could tell the app to quit by writing  to the file.
  This would work on Win too.
  (I didn't add this, but will on request; it's a two-line change)

- didn't include the "in_critical_section > 0" changes; not needed.

David K, can you take a look at this and see if it's OK?

Thanks -- David


___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David E Kim]

I’ll can test this.  That’s the only way I’d know if it’s ok.

It will be easier for me to test if you include the text file messaging if you 
don’t mind/ or I’d just re-implement the signals.

Thanks David!

David K




> On Apr 7, 2017, at 5:13 PM, David Anderson  wrote:
> 
> I made some revisions to David Kim's code and put it in a branch:
> https://github.com/BOINC/boinc/tree/android_api_dpa
> 
> The main changes:
> 
> - I added and fixed comments about the thread/signal structure
> 
> - In exit_from_timer_thread():
>  removed the "boinc_status.suspended = false" and the first "sleep(2.0)".
>  I don't see why either of these is needed.
> 
> - in timer_handler():
>  If detect heartbeat failure, don't bother sending signal to work thread;
>  exit_from_timer_thread() already does this.
> 
> - in worker_signal_handler, don't bother checking that we're the timer thread.
>  Now that we use pthread_kill() instead of setitimer(),this will always be 
> the case.
> 
> - I didn't include the stuff about debugging using external signals.
>  That's a good idea, but there's a simpler way to do it:
>  have the API look for control messages in a file instead of shared memory.
>  e.g. you could tell the app to quit by writing  to the file.
>  This would work on Win too.
>  (I didn't add this, but will on request; it's a two-line change)
> 
> - didn't include the "in_critical_section > 0" changes; not needed.
> 
> David K, can you take a look at this and see if it's OK?
> 
> Thanks -- David

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David Anderson]

I made some revisions to David Kim's code and put it in a branch:
https://github.com/BOINC/boinc/tree/android_api_dpa

The main changes:

- I added and fixed comments about the thread/signal structure

- In exit_from_timer_thread():
  removed the "boinc_status.suspended = false" and the first "sleep(2.0)".
  I don't see why either of these is needed.

- in timer_handler():
  If detect heartbeat failure, don't bother sending signal to work thread;
  exit_from_timer_thread() already does this.

- in worker_signal_handler, don't bother checking that we're the timer thread.
  Now that we use pthread_kill() instead of setitimer(),this will always be the 
case.

- I didn't include the stuff about debugging using external signals.
  That's a good idea, but there's a simpler way to do it:
  have the API look for control messages in a file instead of shared memory.
  e.g. you could tell the app to quit by writing  to the file.
  This would work on Win too.
  (I didn't add this, but will on request; it's a two-line change)

- didn't include the "in_critical_section > 0" changes; not needed.

David K, can you take a look at this and see if it's OK?

Thanks -- David
___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David E Kim]

I made a pull request.  Your instructions worked perfectly, thank you!

David K


> On Apr 7, 2017, at 12:42 AM, Oliver Bock  wrote:
> 
> On 07.04.17 1:24 , David E Kim wrote:
>> How can I push my changes?  I have them ready to push but don’t have
>> permission.
> 
> You don't need any permission. The usual GitHub workflow looks like this:
> 
> - Create an account on GitHub and upload a ssh public key
> - Fork BOINC on GitHub (button in upper-right corner)
> - Clone your fork of BOINC via ssh
> - Create a branch locally (git checkout -b fix-android-api)
> - Make your changes and commit them
> - Push your changes (git push -u origin fix-android-api)
> - Create a pull-request on GitHub (visit your fork, it will offer it)
> 
> Or: https://help.github.com/articles/fork-a-repo
> 
> 
> Let me know if you've got further questions.
> 
> Oliver
> 
> 

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David Kim]

Thanks for this tutorial.  I’ll give it a try tomorrow.

DK


> On Apr 7, 2017, at 12:42 AM, Oliver Bock  wrote:
> 
> On 07.04.17 1:24 , David E Kim wrote:
>> How can I push my changes?  I have them ready to push but don’t have
>> permission.
> 
> You don't need any permission. The usual GitHub workflow looks like this:
> 
> - Create an account on GitHub and upload a ssh public key
> - Fork BOINC on GitHub (button in upper-right corner)
> - Clone your fork of BOINC via ssh
> - Create a branch locally (git checkout -b fix-android-api)
> - Make your changes and commit them
> - Push your changes (git push -u origin fix-android-api)
> - Create a pull-request on GitHub (visit your fork, it will offer it)
> 
> Or: https://help.github.com/articles/fork-a-repo
> 
> 
> Let me know if you've got further questions.
> 
> Oliver
> 
> 

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Oliver Bock]

On 07.04.17 1:24 , David E Kim wrote:
> How can I push my changes?  I have them ready to push but don’t have
> permission.

You don't need any permission. The usual GitHub workflow looks like this:

- Create an account on GitHub and upload a ssh public key
- Fork BOINC on GitHub (button in upper-right corner)
- Clone your fork of BOINC via ssh
- Create a branch locally (git checkout -b fix-android-api)
- Make your changes and commit them
- Push your changes (git push -u origin fix-android-api)
- Create a pull-request on GitHub (visit your fork, it will offer it)

Or: https://help.github.com/articles/fork-a-repo


Let me know if you've got further questions.

Oliver


___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Oliver Bock]

On 06.04.17 17:33 , David Kim wrote:
> but I can definitely push our
> changes to have them reviewed, updated, and tested.

That was the idea :-)
___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Christian Beer]

On 07.04.2017 01:24, David E Kim wrote:
> How can I push my changes?  I have them ready to push but don’t have 
> permission.
>
> It’s just 3 file updates:
>
> api/boinc_api.cpp
> lib/diagnostics.cpp
> lib/diagnostics.h
>
> thanks!

Hi David,

the workflow regarding pull requests is that you need to create a fork
of the BOINC/boinc repository with your github account (davidekim/boinc)
where you push your changes to. Then you can create a pull request (via
the website) from your repository to the main BOINC repository.

Regards
Christian

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David E Kim]

How can I push my changes?  I have them ready to push but don’t have permission.

It’s just 3 file updates:

api/boinc_api.cpp
lib/diagnostics.cpp
lib/diagnostics.h

thanks!

David K


> On Apr 6, 2017, at 5:35 AM, Oliver Bock  wrote:
> 
> If you find the time, it would be appreciated if you'd even push the
> current changes as well :-)
> 
> Thanks,
> Oliver

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David Kim]

I just don't want to break anything out of my lack of familiarity with the 
inner workings of the API but I can definitely push our changes to have them 
reviewed, updated, and tested. 

Sent from my iPhone

> On Apr 6, 2017, at 5:35 AM, Oliver Bock  wrote:
> 
>> On 06.04.17 9:38 , David Kim wrote:
>> Ok, I just posted an issue on GitHub.
> 
> Thanks.
> 
>> I also gave David Anderson our boinc_api.cpp mods.
> 
> For future contributions/fixes it would be great if you'd just push your
> changes to GitHub and open a PR. This way your patches don't leave the
> toolchain, attribution is done correctly and potential discussions are
> done visibly and closely to the code itself.
> 
> If you find the time, it would be appreciated if you'd even push the
> current changes as well :-)
> 
> Thanks,
> Oliver
___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Oliver Bock]

On 06.04.17 9:38 , David Kim wrote:
> Ok, I just posted an issue on GitHub.

Thanks.

> I also gave David Anderson our boinc_api.cpp mods.

For future contributions/fixes it would be great if you'd just push your
changes to GitHub and open a PR. This way your patches don't leave the
toolchain, attribution is done correctly and potential discussions are
done visibly and closely to the code itself.

If you find the time, it would be appreciated if you'd even push the
current changes as well :-)

Thanks,
Oliver
___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David Kim]

I believe it is and will be very much worth the extra work to keep up with 
android but hopefully forward compatibility will be maintained to keep it easy.

David K




> On Apr 6, 2017, at 12:30 AM, Oliver Bock  wrote:
> 
> On 06.04.17 9:25 , Oliver Bock wrote:
>> On 06.04.17 8:11 , Eric Driver wrote:
>>> As a side note, do you get enough android users to justify the extra
>>> work of maintaining the android app?
>> 
>> Luckily BOINC is a community project nowadays so "extra work" is a term
>> that doesn't really fit anymore, as long as there's a community willing
>> to support something - regardless of whether some metric of worthiness
>> applies or not :-)
> 
> That said, the *only* remaining bit that's not yet in the community's
> hand, and thus requires "official" BOINC support, is the Play Store
> account to publish the app. Maybe we can even find a way around that,
> without resorting to only publish via F-Droid...
> 
> Anyhow, any further discussion should probably continue here:
> https://github.com/BOINC/boinc/issues/1846
> 
> Oliver
> 

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David Kim]

Ok, I just posted an issue on GitHub.  I also gave David Anderson our 
boinc_api.cpp mods.

regards,

David K

> On Apr 6, 2017, at 12:19 AM, Oliver Bock  wrote:
> 
> Hi David,
> 
> On 05.04.17 20:39 , David E Kim wrote:
>> I’m not sure if this is the right list to use but thought I’d mention
>> it here.
> 
> If you don't mind, please open an issue for this on GitHub to facilitate
> tracking down this problem.
> 
> Thanks,
> Oliver

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Oliver Bock]

On 06.04.17 9:25 , Oliver Bock wrote:
> On 06.04.17 8:11 , Eric Driver wrote:
>> As a side note, do you get enough android users to justify the extra
>> work of maintaining the android app?
> 
> Luckily BOINC is a community project nowadays so "extra work" is a term
> that doesn't really fit anymore, as long as there's a community willing
> to support something - regardless of whether some metric of worthiness
> applies or not :-)

That said, the *only* remaining bit that's not yet in the community's
hand, and thus requires "official" BOINC support, is the Play Store
account to publish the app. Maybe we can even find a way around that,
without resorting to only publish via F-Droid...

Anyhow, any further discussion should probably continue here:
https://github.com/BOINC/boinc/issues/1846

Oliver

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Oliver Bock]

On 06.04.17 8:11 , Eric Driver wrote:
> As a side note, do you get enough android users to justify the extra
> work of maintaining the android app?

Luckily BOINC is a community project nowadays so "extra work" is a term
that doesn't really fit anymore, as long as there's a community willing
to support something - regardless of whether some metric of worthiness
applies or not :-)

Best,
Oliver

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Oliver Bock]

Hi David,

On 05.04.17 20:39 , David E Kim wrote:
> I’m not sure if this is the right list to use but thought I’d mention
> it here.

If you don't mind, please open an issue for this on GitHub to facilitate
tracking down this problem.

Thanks,
Oliver
___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Eric Driver]

Hi David,

A couple years ago I built an android app for the NumberFields@home project.  I 
too saw random seg faults about 50% of the time, and this was with code that 
worked perfectly on linux/windows/mac platforms.  I never figured it out, 
primarily because I was too busy to learn how to use the android debugger.  I 
put it on the back burner and never came back to it.  So I'm glad to see you 
may have figured it out.

As a side note, do you get enough android users to justify the extra work of 
maintaining the android app?  

Thanks,
Eric



-Original Message-
From: boinc_dev [mailto:boinc_dev-boun...@ssl.berkeley.edu] On Behalf Of David 
E Kim
Sent: Wednesday, April 05, 2017 11:40 AM
To: boinc_dev@ssl.berkeley.edu
Subject: [boinc_dev] BOINC API issue with android arm?

Hi all,

I’m not sure if this is the right list to use but thought I’d mention it here.  
 I tried to contact David Anderson via email but have not heard back.

I was wondering if anyone has random seg fault issues with their android apps 
for their BOINC projects.  Our Rosetta@home android application had pretty bad 
success rates and I found that the main cause was the use of “setitimer” in the 
“start_worker_signals” function in the boinc_api.cpp code.  Our Rosetta app 
runs stable without the BOINC api, but when using the default api, it crashed 
randomly.

I developed a workaround which avoids “setitimer" in the api code and our app 
is stable with the modified BOINC api and is now in production for our 
Rosetta@home project.  Our success rates for android arm have gone from 40-60% 
to 95%.

I’m wondering if this is just a Rosetta application induced bug (it’s a pretty 
substantial scientific app) or a general android issue.

Thanks,

David Kim
___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David Anderson]

I'll look at this now; I was swamped earlier.
-- David

On 4/5/2017 11:39 AM, David E Kim wrote:

Hi all,

I’m not sure if this is the right list to use but thought I’d mention it here.  
 I tried to contact David Anderson via email but have not heard back.

I was wondering if anyone has random seg fault issues with their android apps 
for their BOINC projects.  Our Rosetta@home android application had pretty bad 
success rates and I found that the main cause was the use of “setitimer” in the 
“start_worker_signals” function in the boinc_api.cpp code.  Our Rosetta app 
runs stable without the BOINC api, but when using the default api, it crashed 
randomly.

I developed a workaround which avoids “setitimer" in the api code and our app 
is stable with the modified BOINC api and is now in production for our Rosetta@home 
project.  Our success rates for android arm have gone from 40-60% to 95%.

I’m wondering if this is just a Rosetta application induced bug (it’s a pretty 
substantial scientific app) or a general android issue.

Thanks,

David Kim
___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.


___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[David E Kim]

Thanks for the feedback.

Rosetta@home for Android arm is stable now.  It’s built, tested, and stable for 
Android 4 to 7.+ (nougat) versions.  I’m going to test it for beta “Oreo" as 
soon as we get our test device.  If this is a general problem, then I can show 
whoever is interested in what I did.

I wonder why some projects are stable and others are not.  There is definitely 
an issue with the use of “setitimer”.

David K


> On Apr 5, 2017, at 11:46 AM, Vitalii Koshura  
> wrote:
> 
> Hello David,
> 
> I'm an active boinc android user and can say that I see such behavior on 
> several projects (including Rosetta). Also as I've noticed that app crashes 
> when boinc pausing applications (in this case all running applications are 
> marked as failed). Other science applications (e.g. SETI@home, World 
> Community Grid) have no such problem.
> 
> I hope this can help you to understand what's going wrong.
> 
> Thanks
> 
> Best regards,
> Vitalii Koshura
> 
> 2017-04-05 21:39 GMT+03:00 David E Kim  >:
> Hi all,
> 
> I’m not sure if this is the right list to use but thought I’d mention it 
> here.   I tried to contact David Anderson via email but have not heard back.
> 
> I was wondering if anyone has random seg fault issues with their android apps 
> for their BOINC projects.  Our Rosetta@home android application had pretty 
> bad success rates and I found that the main cause was the use of “setitimer” 
> in the “start_worker_signals” function in the boinc_api.cpp code.  Our 
> Rosetta app runs stable without the BOINC api, but when using the default 
> api, it crashed randomly.
> 
> I developed a workaround which avoids “setitimer" in the api code and our app 
> is stable with the modified BOINC api and is now in production for our 
> Rosetta@home project.  Our success rates for android arm have gone from 
> 40-60% to 95%.
> 
> I’m wondering if this is just a Rosetta application induced bug (it’s a 
> pretty substantial scientific app) or a general android issue.
> 
> Thanks,
> 
> David Kim
> ___
> boinc_dev mailing list
> boinc_dev@ssl.berkeley.edu 
> https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev 
> 
> To unsubscribe, visit the above URL and
> (near bottom of page) enter your email address.
> 

___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Re: [boinc_dev] BOINC API issue with android arm?

[Vitalii Koshura]

Hello David,

I'm an active boinc android user and can say that I see such behavior on
several projects (including Rosetta). Also as I've noticed that app crashes
when boinc pausing applications (in this case all running applications are
marked as failed). Other science applications (e.g. SETI@home, World
Community Grid) have no such problem.

I hope this can help you to understand what's going wrong.

Thanks

Best regards,
Vitalii Koshura

2017-04-05 21:39 GMT+03:00 David E Kim :

> Hi all,
>
> I’m not sure if this is the right list to use but thought I’d mention it
> here.   I tried to contact David Anderson via email but have not heard back.
>
> I was wondering if anyone has random seg fault issues with their android
> apps for their BOINC projects.  Our Rosetta@home android application had
> pretty bad success rates and I found that the main cause was the use of
> “setitimer” in the “start_worker_signals” function in the boinc_api.cpp
> code.  Our Rosetta app runs stable without the BOINC api, but when using
> the default api, it crashed randomly.
>
> I developed a workaround which avoids “setitimer" in the api code and our
> app is stable with the modified BOINC api and is now in production for our
> Rosetta@home project.  Our success rates for android arm have gone from
> 40-60% to 95%.
>
> I’m wondering if this is just a Rosetta application induced bug (it’s a
> pretty substantial scientific app) or a general android issue.
>
> Thanks,
>
> David Kim
> ___
> boinc_dev mailing list
> boinc_dev@ssl.berkeley.edu
> https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> To unsubscribe, visit the above URL and
> (near bottom of page) enter your email address.
___
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.