id is appended after the name instead of after the extension. Avoid problems
with the VirtualBox interface.
samples/vboxwrapper/
vbox.cpp, .h
vboxwrapper.cpp, .h
svn path=/trunk/boinc/; revision=24880
environment and they suspend/resume the VM by the VirtualBox application.
samples/vboxwrapper/
vbox.cpp, .h
vboxwrapper.cpp
svn path=/trunk/boinc/; revision=24878
for virtual hard disks.
- VBOX: Execute closemedium against the virtual floppy disks when the VM is being
deleted.
samples/vboxwrapper/
vbox.cpp
vboxwrapper.cpp
svn path=/trunk/boinc/; revision=24877
and restarted it would null out the host port information.
- VBOX: Try to make the CERN compatible datatype for FloppyIO more JSON friendly
by following the advice of issue:
https://github.com/stig/json-framework/issues/48
- VBOX: Support both the old style and new style of determining assigned threads
of execution (aka core counts). 6.12 and the early versions of 7.x support
the old style.
samples/vboxwrapper/
vbox.cpp, .h
vboxwrapper.cpp, .h
svn path=/trunk/boinc/; revision=24876
a bug that was fixed last night which prevented the wrapper from properly
reporting its checkpoint time to the CC. The lack of checkpoint information
for the job would have prevented the CC from rescheduling the job and removing
it from memory. Now it can so the extra shutdown each scheduling period
should not be needed.
samples/vboxwrapper/
vboxwrapper.cpp
svn path=/trunk/boinc/; revision=24874
VboxSVC.exe is normally launched by svchost.exe which only contains environment
variables defined by the system. In the BOINC sandbox environment we want to
add VBOX_USER_HOME to redirect reading/writing configuration files to a
different directory. What happens if we launch it ahead of making any calls to
VBoxManage which would cause the OS to launch it? If we launch it, it should
inherit our environment block and work just like it works on Linux and the Mac.
This is such a simple solution, why didn't I think of it sooner?
samples/vboxwrapper/
vbox.cpp
svn path=/trunk/boinc/; revision=24873
rules for a VM until the VM is in a powered off state. I guess the VM will
just have to fail starting up if it cannot allocate the assigned port.
samples/vboxwrapper/
vbox.cpp
svn path=/trunk/boinc/; revision=24863
state, deregistering the VM would fail all commands because the VM was not
mutable. The first thing we should do when deregistering is discard
any saved state.
samples/vboxwrapper/
vbox.cpp
svn path=/trunk/boinc/; revision=24862
then all the deregister commands have to be optional. Go ahead and log errors
to stderr but continue to run through the various commands anyway.
samples/vboxwrapper/
vbox.cpp
svn path=/trunk/boinc/; revision=24860
vboxwrapper. Not doing so leaves the drive reference around and the VM in a
saved state. The next attempt to delete it from another instance of
vboxwrapper fails with a complaint about the old unnamed vm(all files have
been cleaned up by BOINC) being in an non-mutable state.
samples/vboxwrapper/
vbox.cpp
vboxwrapper.cpp
svn path=/trunk/boinc/; revision=24859
are looking at the existing state of the host.
- VBOX: Reset the output buffer for every command execution.
- VBOX: Remove the teleporting and snapshotting states from the list of running
states.
- VBOX: Remove the '--type floppy' typo when mounting the floppy disk image.
samples/vboxwrapper/
vbox.cpp, .h
svn path=/trunk/boinc/; revision=24858
Some more work will be required if there is ever a need for more than
one port. But this should work for now. This should successfully
handle running multiple instances of the same job, port information is
stored in a file called "vbox_firewall.txt" which is updated for each
execution of the wrapper.
samples/vboxwrapper/
vbox.cpp, .h
vboxwrapper.cpp
win_build/
vboxwrapper.vcproj
svn path=/trunk/boinc/; revision=24843
scale their PFC by 0.1 in credit calculations.
This reflects the fact that GPU apps are typically less efficient
(relative to device peak FLOPS) than are CPU apps.
The actual values from SETI@home and Milkyway are 0.05 and 0.08.
svn path=/trunk/boinc/; revision=24842
wall clock time a VM is allowed, if exceeded shutdown gracefully and tell
BOINC that the job is finished.
samples/vboxwrapper/
vbox.cpp, .h
vboxwrapper.cpp
svn path=/trunk/boinc/; revision=24837
to an account manager via cookies. This fix is good enough for the 6.8 client.
We'll need to do a little more for 7.x and beyond.
clientgui/
WizardAttach.cpp
svn path=/trunk/boinc/; revision=24827
to a superceded or deprecated app version, use it anyway.
The current app version may not validate against the old one.
svn path=/trunk/boinc/; revision=24823
registered with VirtualBox during the RegisterVM phase. If we detect the
already exists error message run the unsupported internal command to reset
the HD UUID which then allows us to attach the HD to the VM. This should allow
us to run multiple jobs with the same base VM HD image.
samples/vboxwrapper/
vbox.cpp
svn path=/trunk/boinc/; revision=24822
- client: msg tweak
- client: minimum work buffer lower bound is 180 sec
- scheduler: in computing HOST_USAGE::project_flops for a job,
if we don't have sufficient elapsed_time statistics
for either the (host, app_version) or the app_version,
use a conservative estimate (p_fpops*(#cpus+#ngpus))
rather than the number returned by app_plan().
This avoids "time limit exceeded" errors when the latter is way off.
svn path=/trunk/boinc/; revision=24820