The gnulib 'error' module uses 'program_name'. On some platforms --
but not Linux / glibc -- it references it as:
extern char *program_name;
This means when you compile libguestfs on non-glibc (eg. Mac OS X)
gnulib requires 'program_name' as an external string reference, which
we don't provide.
This change doesn't define this string reference for gnulib, but it
does change the name of the macro we use to avoid conflicts if we
eventually need to export 'program_name' as a string.
Thanks: Margaret Lewicka
When copying disks that use EFI, we created a new partition table,
randomizing the GPT GUID of the first partition. Since EFI may store
the GUID in its NVRAM variables, this could make the guest unbootable.
In GPT, each partition has a GUID assigned randomly. Allow this GUID
to be written and read.
Note this is different from the GUID type code which is used to
identify the type of the partition.
Commit 9135129b0f changed
two stack buffers to pointers:
- uint8_t buf[BUFFER_SIZE];
- unsigned char outbuf[BUFFER_SIZE];
+ CLEANUP_FREE uint8_t *buf = NULL;
+ CLEANUP_FREE uint8_t *outbuf = NULL;
but we were still using sizeof buf to calculate the size of the
buffer. sizeof buf == 8 so the original code which used large buffers
for reading/writing the file changed to using 8 byte buffers.
Currently implemented as guestfish commands, provide them instead as
single source -> destination functions for the library, so they can be
used also in other places.
These functions are not added to guestfish, since guestfish has its own
implementation (which will soon switch to call copy-in and copy-out for
multiple paths).
test-dlopen needs to be added only when libdl is available, and it is
like that already (within if HAVE_LIBDL); hence don't unconditionally
add it as well.
Check whether --part-type is supported (in the rewritten sfdisk in
util-linux >= 2.26), and use it instead of --print-id & --change-id.
The actual result should be the same, just not using a (recently)
deprecated command line API.
Unlike -o rhev which have only one data domin, -o vdsm can and usually
does have multiple data domain.
The path to vdsm is pre mounted so no need to assume nfs path with -os
Example:
-o vdsm -os /rhev/data-center/<data-center-uuid>/<data-domain-uuid>
Bug-Url: https://bugzilla.redhat.com/1176591
Signed-off-by: Shahar Havivi <shaharh@redhat.com>
Kernel 3.19 has problems uncompressing the RAM disk with <= 500 MB.
(This is likely to be a kernel bug)
64 KB pages are common on aarch64, so treat this case the same as ppc,
and use a larger default appliance memory size.
Thanks: Laszlo Ersek for help and reproducing the bug.
Since libguestfs >= 1.26 / supermin 5, the appliance has lived in a
subdirectory of $TMPDIR/.guestfs-$UID. As a result using 'ls -l'
didn't display any information about the files in the appliance.
Using '-R' causes this information to be displayed again.
ls -Z stops ls from displaying the size field. --lcontext uses longer
lines but has the advantage of still displaying the size field as well
as the SELinux context.
It is documented as 'p2v', but in fact this was not implemented
so there is no way to log in as root on the console.
On RHEL we also have to add the 'passwd' package, else you get:
Error creating Live CD : Unable to set unencrypted password due to lack of /usr/bin/passwd
Covert common_utils_tests to use oUnit as testing framework, replacing
the hand-made assert in favour of structured unit tests and better
error reporting.
common_utils_tests is now built only when the oUnit module has been
found.
Pass to --commands-from-file the name of a file containing customization
commands in each line, as if they were specified as command line
arguments.
This eases the reuse of commands among different
builder/customize/sysprep invocations.
