Split in an own function the code dong the mounting, the inspection of
the kernel files, and the downloading, including a per-OS function for
the actual kernel files. This gives few advantages:
- the download phease is not repeated for all the files to fetch
- it is easier to eventually support multi-root disk images
- it is possible to support OSes different than Linux; virt-get-kernel
now will just report the unsupported OS, instead of a generic
"no kernel found" message
This is mostly code refactoring, with (on Linux) no actual behaviour
change.
Make print_version_and_exit, long_options, display_short_options, and
display_long_options private, as set_standard_options now takes care of
handling the job for the common command line options.
Move --debug-gc as common option for all the OCaml-based tools, even a
couple of them which didn't have it previously.
As implementation note, make set_debug_gc private to
set_standard_options, as it needed to be moved otherwise, and it is no
more required as public function.
Introduce a new common helper to add the common options for libguestfs
tools (short/long options, version, verbose, trace), and sort them.
All the OCaml-based tools had these options already, so there are no
functional changes in the interface they provide.
The only difference is that now the options are always sorted, while
some tools didn't had them like that previously: because of this, a
couple of ditto markers (as descriptions) don't match what's above them
anymore, and thus their full description is put instead.
When running (eg) dnf on a 32 bit i686 guest when the host is 64 bit
x86_64, dnf believes it is running on a 64 bit machine and so tries to
install x86_64 packages. We can 'trick' dnf into believing it's a 32
bit machine using the setarch program.
$ virt-builder fedora-22 --arch i686 --install 'gperf'
...
[ 27.4] Installing packages: gperf
...
Running transaction test
Error: Transaction check error:
package libgcc-5.1.1-4.fc22.x86_64 is intended for a different architecture
...
The use of a heredoc to solve quoting issues comes from:
http://stackoverflow.com/a/3435460
Thanks: Jan Sedlák for finding the solution.
In -i libvirtxml, -i ova and -o libvirt drivers, replace the ad hoc
xpath_to_* functions with use of the new xpath convenience functions
introduced in the previous commit.
This is not entirely refactoring because I fixed a few bugs found by
type safety.
Previously given a path such as:
vpx://vcenter/Folder/Datacenter/esxi
we calculated dcPath=Folder. However this is obviously wrong. We
should chop the path at the final (esxi) element to give
dcPath=Folder/Datacenter.
In vCenter, the datacenter is identified by a path, and the parameter
used for this is called 'dcPath'. Rename the function to avoid any
confusion about what we're getting here.
This is just renaming function/variable names and has no other effect.
Several tools handle --debug-gc command-line option, by explicitly
forcing GC on every exit path. This is tedious and prone to forgetting
some of the exit paths.
Instead, add a generic handler for --debug-gc, which installs an at_exit
hook to do the GC consistency check, and which can be called right in
the command-line parser. Also adjust all users of --debug-gc to use
that handler.
Signed-off-by: Roman Kagan <rkagan@virtuozzo.com>
Build the P2V disk image and boot it. We don't actually use PXE
specifically, but we do test the whole PXE / kernel command line path
much more thoroughly.
This is a 'check-slow' test because it takes ages to run.
When using the virt-p2v ISO in command line mode, we did not wait for
the network to come online before starting virt-p2v. Therefore
virt-p2v could exit with an error when testing the ssh connection (or
on the other hand, it might work randomly). If the user logs in and
runs 'launch-virt-p2v' by hand, then it would usually work because the
network had been brought online in the meantime.
Fix this by waiting for NetworkManager to bring the connection online
before calling test_connection(). Note that the obvious way to fix
this (changing the systemd service to wait for network-online.target)
does *not* work - I added a comment to the service about this.
Thanks: Tingting Zheng
Raise a warning for each kernel specified in grub which does not
actually exist, keep going on with the conversion using the remaining
(existing) kernels.
Put in a list the errnos to expose, filling the content of the
Guestfs.Errno submodule from that.
Also, generate a separate guestfs-c-errnos.c with the implementations of
the functions returning the errno codes.
Only code motion and refactoring, no actual changes on the content of
the ocaml Guestfs module.
pcre < 8 has limitations on the types of partial matching
possible. As noted in the pcrepartial(3) man page, these can
be worked around by adding non-capturing parentheses around
various subexpressions, as I have done here.
Apparently gtk_init_check doesn't (always?) print an error message
when it returns false, so virt-p2v just exits without any message.
This was observed on RHEL 6 when $DISPLAY was not set.
File input_libvirt_other.ml, line 1, characters 0-1:
Error: The implementation input_libvirt_other.ml
does not match the interface input_libvirt_other.cmi:
Class declarations do not match:
class virtual input_libvirt :
'a ->
string option ->
string ->
object
method adjust_overlay_parameters : Types.overlay -> unit
method as_options : string
method virtual source : unit -> Types.source
end
does not match
class virtual input_libvirt :
string option ->
string option ->
string ->
object
method adjust_overlay_parameters : Types.overlay -> unit
method as_options : string
method virtual source : unit -> Types.source
end
A parameter has type 'a but is expected to have type string option
When the output format is qcow2 and -oa preallocated is used,
previously it would only allocate the metadata. That was a regression
over previous behaviour of virt-v2v 0.9. Change it so it allocates
the full file size in this case.
For raw, this allows "off" as a synonym for "sparse" (to make it
consistent with qcow2).
For qcow2, this allows "sparse" as a synonym for "off".
It also adds qcow2 "full" preallocation, which is actually mapped to
the qemu option "falloc" (see arguments about this on the qemu-devel
mailing list, which we lost).
This also updates the test.
Normalize the target architecture, and also each architecture when
checking for a compatible image.
This sort of reverts the effects of
commit 8864c47b94, but at least it is
possible to build e.g. Debian-based amd64 images on any x86_64 system
without being considered as foreign architecture.
Small helper to normalize an architecture string, so it is easier to
compare them and check for compatibilities.
Make use of it in guest_arch_compatible, simplifying it.
If a function name, with its trailing parentheses, is in the
environment , trying to unset it will error out with a message of "not
a valid identifier". Handle this as a special case, and unset it with
the -f option which can handle the parentheses in the supplied
identifier.
