When TESTS_ENVIRONMENT already uses 'run', the VG variable
doesn't also need to use 'run'.
The specific problem is that if the command contains newlines
then double invocations of the 'run' script fails (in libtool).
ie the following command failed causing errors in check-valgrind:
$VG virt-builder phony-fedora \
-v --no-cache --no-check-signature $no_network \
...
--write '/etc/append4:line1
' \
...
Currently 'make install' installs the virt-p2v binary in
/usr/libexec/virt-p2v on the host. It is never supposed to be run
from there, even by another program, so use of /usr/libexec is
incorrect. It is only supposed to be copied into USB keys / ISOs /
etc created by virt-p2v-make-* scripts.
The other problem with shipping a "naked" binary on the host is that
packages built from that get all the dependencies of virt-p2v, for
example Gtk. This is unnecessary just for running the command line
scripts mentioned above.
This changes the Makefile and scripts so that the binary is stored
compressed in $libdir/virt-p2v/virt-p2v.xz. It is compressed to avoid
exposing the dependencies. It is stored under $libdir since the
binary is still architecture-dependent.
A further change is that when we copy the binary into the virt-p2v
ISO, it is now installed in /usr/bin instead of /usr/libexec. (And
note that we always use /usr/bin, not $bindir, since this path should
not need to be affected by the configuration of libguestfs).
Added:
- cdrtools: added as alternative to cdrkit
- multipath-tools: contains kpartx (in AUR)
Removed:
- ntfsprogs: the package is no longer available, it has been completely
replaced by ntfs-3g (already in packagelist.in)
- zfs-fuse: no longer in AUR
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
The OVF standard allows the use of SHA256 hashes in the manifest file.
Adding support for this.
One of the tests was updated to put SHA256 into manifest file.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
The regular expression for parsing the manifest line was wrong. There is
a mandatory space between '=' and the hash.
Another problem was that only the first line of the manifest file was
actually processed.
Also added some debugging info and warning to catch problems with
parsing.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Try to improve the way packages of VMware tools are removed from
YUM-based guests:
- when filtering the package itself from its providers, do a stricter
check so either the provide is the unversioned package, or it is
exactly its own name
- if the package has no other providers, then going further will cause
the invocation of 'yum install' with no packages, and thus the package
itself will not be added to the list of packages to be removed; to
overcome this issue, just mark the package as "to be removed" in that
case
Related to: RHBZ#1155150
When checking whether a kernel supports virtio or it is Xen-based, it is
assumed that the feature has the kernel module for it; this will fail if
the feature is built-in in the kernel, misrepresenting it.
The solution is to check the kernel configuration whether the feature
is built-in, in case there is no module available.
This fixes the virtio detection on Ubuntu kernels, where virtio is
built in and not as module.
Detect only once which method must be used to get and set the default
Grub2 kernel in the guest: this avoids the same checks in list_kernels
and set_default_kernel.
Also, add a "no method" option as well: Debian/Ubuntu guests do not have
neither grubby nor Perl's Bootloader::Tools, so there is no way to know
what is the default kernel, nor to change it. In this case, add a
warning about this situation.
Move the Checksums module from virt-builder mostly as it is; the only
change is that on checksum mismatch an exception is raised rather than
invoking "error" directly: this way users of verify_checksum &
verify_checksums can do their own handling of the situation.
Debian is also doing a UsrMove/UsrMerge:
https://wiki.debian.org/UsrMerge
However it is not finalized that Debian will actually make this
change.
Since some Debian systems have /sbin as a symlink and other have /sbin
as a real directory, and we should avoid choosing a symlinked
directory for the daemon, the easiest fix is simply to probe /usr/sbin
before /sbin since under all scenarios (and Fedora too) /usr/sbin is a
real directory.
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=838995
Create a new libfishcommon convenience static library to build just once
(instead of 12 times!) the majority of the guestfish sources used in the
rest of the C tools (mostly for command line stuff, inspection, and
mount).
The notable exceptions not using libfishcommon are guestfish itself, and
virt-rescue: both need to build at least one of the common sources using
additional CPPFLAGS.
All the define was doing in options.h was masking the declaration of
inspect_mount_root, which was always built-in in inspect.c (because of
the unconditional #define there) anyway.
Since this is common code used by all the C tools, try to avoid extra
knobs which add different code paths for no benefit.
Some of the C tools were building also config.c as part of the shared
sources from guestfish, and thus bringing a dependency on libconfig.
Since none of them actually read the libguestfs configuration at all,
then exclude fish/config.c from their build, and stop linking to
libconfig.
Using update-initramfs is the native way of updating initrd on Debian
based systems.
To add some modules to the image we can list them in file
/etc/initramfs-tools/modules.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
A disk of type 'volume' is stored as
<source pool='pool_name' volume='volume_name'/>
and its real location is inside the 'volume_name', as 'pool_name': in
this case, query libvirt for the actual path of the specified volume in
the specified pool.
Adjust the code so that:
- for_each_disk gets the virConnectPtr, needed to do operations with
libvirt
- when extracting the disk filename depending on the type, the code
snippet doing it can directly set 'filename', without setting an XPath
result variable
Only file-based volumes are supported for now; more types can be added
(with proper testing) later on.
Since libvirt 2.0.0, these two new <listen/> types have been
supported: https://libvirt.org/formatdomain.html#elementsGraphics
This change just copies that configuration over from the source to the
destination if the destination is also libvirt.
Since we previously used 'LNone' to mean "no parseable <listen/>
element" I also had to change previous uses of 'LNone' to 'LNoListen',
so we can use 'LNone' to mean "<listen type='none'>".
Thanks: Ming Xie.
NTFS file system always has the MFT file at inode 0. This reliable
information helps testing the API.
Signed-off-by: Matteo Cafasso <noxdafox@gmail.com>
Library's counterpart of the daemon's internal_find_inode command.
It writes the daemon's command output on a temporary file and parses it,
deserialising the XDR formatted tsk_dirent structs.
It returns to the caller the list of tsk_dirent structs generated by the
internal_find_inode command.
Signed-off-by: Matteo Cafasso <noxdafox@gmail.com>
The internal_find_inode command searches all entries referring to the
given inode and returns a tsk_dirent structure for each of them.
The command is able to retrieve information regarding deleted
or unaccessible files where other commands such as stat or find
would fail.
The gathered list of tsk_dirent structs is serialised into XDR format
and written to a file by the appliance.
Signed-off-by: Matteo Cafasso <noxdafox@gmail.com>
Make use of the additional command line arguments, and API needed to
decrypt LUKS partitions.
This affects only virt-customize, virt-get-kernel, virt-sparsify, and
virt-sysprep, as they are the main OCaml tools interacting with
user-provided images.
Expose via Common_utils the C functions & variables (part of guestfish)
that handle decryption of LUKS partitions, and the additional command
line arguments to tune the way they work. This way it will be easy to
provide (basic) crypto support also in OCaml-based tools.
Related to: RHBZ#1362649
With the current implementation, the root inode of the given partition
is ignored.
The root inode is now reported. Its name will be a single dot '.'
reproducing the TSK API.
Signed-off-by: Matteo Cafasso <noxdafox@gmail.com>
On Debian family of OSes Grub2 tools are prefixed with 'grub-', not with
'grub2-'. We have to detect the correct name of the tool to use it.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
