For optional arguments, you can now specify empty string to mean no
argument, except for String optional arguments where you must use
"NOARG" (empty string meaning a supplied empty string argument).
This is needed because older versions of grub(for example in centos)
do not understand filesystems created with newer version of e2fsprogs.
By default in e2fsprogs 1.4+ creates partitions with 256 bit inode
size, and grub expect 128 bit size.
This API returns the guest's favicon if found, else an icon
representing the guest operating system. Currently supported by this
patch: Fedora, RHEL and derivatives, Debian (but not Ubuntu),
Windows XP, Windows 7.
This also updates virt-inspector to include an <icon> element
containing the icon in base64 encoding.
The previous guestfs___download_to_tmp function did not handle
multiboot correctly. In particular it used the same cache name
for downloaded files from different roots, which could have caused
things like applications in each root to be confused.
This changes the function so that the cache filename is prefixed
with the root / fs number, eg. $tmpdir/0-Name instead of $tmpdir/Name.
This change also requires the function to return the new name, so
all places in the code which called this function had to be
updated.
This updates and fixes commit 3c1f762abe.
No current function returns RBufferOut and has optional args. Such
functions would be generated incorrectly.
RBufferOut implies a silent "size_t *size_r" argument is added after
the regular arguments and before the optional arguments. Various
changes to the code generator need to be made to take this into
account.
This is a fairly pointless note/warning since (a) you can't use the
API if you don't pass a root device string and (b) the code gives you
a good error message if you pass something that is not a root device
string.
Before gcc 4.5, the deprecated option did not take the optional string
argument (see [1]). This caused compilation to fail with gcc < 4.5.
[1] http://gcc.gnu.org/gcc-4.5/changes.html
If the external command failed to run, we could free up the allocated
*stdoutput and *stderror pointers, but then return those freed
pointers to the caller. The caller usually tries to print and free
*stderror, so this is a serious error.
Instead, return *stdoutput as NULL, and *stderror pointing to a
generic error message.
The given timeout (10s) was too low if the appliance was
running slowly, which caused a cascade of other failures
during tests.
Note that in udev-171 and above on Fedora, /sbin/start_udev
no longer exists, so now we are using this manual method to
start udevd.
