Hi Jacob,
I am not a fan of using the word hacked. It suggests corners are being
cut. We are trying to do things correctly. The modified code is in a
fork and the changes were provided as a PR over 4 months ago. We used
our dev and prod projects to test the changes and have been running this
way while we wait for the changes to be included upstream. Until these
changes are included upstream, we will be using the fork.
Cheers,
Laurence
On 06/11/17 14:52, Jacob Klein wrote:
Hi Laurence,
2 things.
1) I usually only do RNA World tasks (which rarely have VMs
available), and set No New Tasks for any other VM project (because I
want to give the RNA World monster task the best chance to complete
successfully). But here, I needed a simple way to get new VM tasks, to
test the new vboxwrapper. So, yes, I do need some VM project to test
with, and I chose LHC. If you can think of a better one to test with,
let me know.
2) Are you arguing semantics? If an app isn't using an official
version of the vboxwrapper for an app, then it is using a hacked
version. ie: The user has no idea what could be in that executable. If
you are really making fixes to the codebase, then you should please
consider making sure those fixes land in the official codebase!
Development may be skeleton on it currently, but we are obviously
still doing releases. And if your changes are in the codebase already,
then you might be able to switch to using official 26200 soon.
My research indicates that 2 of the 4 projects I tested with, are
indeed currently using hacked versions.
Official 26198 x64: vboxwrapper_26198_windows_x86_64.exe --- Size:
1,375,232
LHCb version: vboxwrapper_26198ab7_windows_x86_64.exe --- Size: 1,696,768
Theory version: vboxwrapper_26198ab5_windows_x86_64.exe --- Size:
1,695,744
Kind Regards,
Jacob
------------------------------------------------------------------------
*From:* Laurence <laurence.fi...@cern.ch>
*Sent:* Monday, November 6, 2017 3:17 AM
*To:* Jacob Klein; David Anderson; Juha Sointusalo; Rom Walton
*Cc:* BOINC Developers Mailing List; Boinc Projects; BOINC Alpha
*Subject:* Re: [boinc_projects] [boinc_dev] BOINC - October 2017
VirtualBox and VBoxWrapper updates
Hi Jacob,
On 05/11/17 10:14, Jacob Klein wrote:
David / Juha / Rom:
I have concluded my testing (took about 6 hours), and 26200 has some serious
problems!
Attached is a .zip of my full results, across my testing with LHC@Home and RNA
World. It includes a results summary, results detail files of stderr.txt logs,
and the app_info.xml files I used.
You don't need to test on LHC@home as are building our own vboxwrapper
while we are waiting for PR 1946
<https://github.com/BOINC/boinc/pull/1946> to be merged.
A summary of the issues is below. Please let me know when you think you have
these issues fixed, and when there is a newer version for me to test!
Thanks,
Jacob
The main issues with 26200 are:
- (Major) The version number written to stderr.txt is incorrect! I believe it reads the
file info, and for 26200, is printing: "vboxwrapper (7.9.26197)". This needs to
be fixed, it is very confusing when testing/debugging!
- (Major) For all 4 of the LHC@Home VM projects, when trying to start a new
task using 26200 using VirtualBox v5.2.x, you immediately get Computation Error.
...
stderr.txt says:
2017-11-05 03:20:47 (12552): Error 0x80bb0001 in vbox52::VBOX_VM::create_vm
(c:\users\david\documents\boinc_git\boinc\samples\vboxwrapper\vbox_mscom_impl.cpp:835)
2017-11-05 03:20:47 (12552): Error Source : SessionMachine
2017-11-05 03:20:47 (12552): Error Description: This machine does not have any
snapshots
Unhandled Exception Detected...
- Unhandled Exception Record -
Reason: Access Violation (0xc0000005) at address 0x00007FFAEAC4AB7F read
attempt to address 0x00000018
- (Major) For RNA World, when trying to restore a task using 26200 using
VirtualBox v5.2.x (which I attempted using a copy of my data while offline!)..
it fails to restore properly! :(
...
stderr.txt says:
2017-11-05 01:37:51 (14600): Restore from previously saved snapshot.
Unhandled Exception Detected...
- Unhandled Exception Record -
Reason: Access Violation (0xc0000005) at address 0x00007FFAEA5E8246 write
attempt to address 0xE8600000
- (Minor) For LHCb and Theory, when trying to start a new task using 26200
using VirtualBox v5.1.x, they will eventually fail mid-run. I'm not entirely
sure, but this might be because they use a hacked VBoxWrapper for these 2 apps,
and a 'standard' VBoxWrapper just won't work.
We haven't "/hacked/" the vboxwrapper, we spent 2 months debugging it!
The issues you hit is that port 9618 is blocked for out going traffic
on your machine.
http://lhcathome.web.cern.ch/test4theory/my-firewall-complaining-which-ports-does-project-use
Cheers,
Laurence
_______________________________________________
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.
