GitLab

have to be shelved soon.

in the sliver so that os_setup does not wait until timeout.  Also some
bug fixes and cleanup all through start/wait for sliver.

Probably still not complete, but closer.

that the pool daemon does not run while the mapper is working.
This will only affect experiments that are using shared nodes
though, since the lock is way too coarse and basically serializes
the mapper. Since very experiments use hared nodes at this time, it
will not really be a problem.

Also added TERM handling to avoid killing the daemon while its
working, as when stopping testbed daemons to update the installation.

update file in each of the sql and install update dirs.

Try to use as much memory as we can for frisbee's buffers.  The amount
we try to allocate is based on the amount of installed memory, so it
will vary from system to system.

Got comparison backwards when checking file date against
HTTP_IF_MODIFIED_SINCE in fetch-vis.