| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
|
|
|
|
|
|
|
| |
We've got this shiny new method selection UI but if you want to use it, you've
first got to sit through the delay of fetching metadata for the closest mirror
which might be completely wrong for you. Thus, this parameter will skip the
default action. Entering and leaving the source spoke will unset this
parameter.
|
| |
|
| |
|
|
|
|
| |
Merge baseudev into storage.udev.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is the first pass at setting things up to split storage out into
a separate package.
It replaces the guts of iutil.execWithRedirect and iutil.execWithCapture
with a single backend function, _run_program. The main thing I can think
of that might be cause for complaint is that _run_program does not log
each output line as the external program outputs it, so you get the whole
output buffer at once. Storage gets its own copy of _run_program and
all of the calls in storage use the copy from storage.util.
The reason I moved all of the iutil arch functions into storage is that
they are mostly used in storage and I didn't want there to be two copies
of the whole set.
The rest is removing of unused things and moving of things only used by
storage into storage.
|
| |
|
|
|
|
| |
This will raise an exception if the thread quit because of an error.
|
| |
|
|
|
|
|
|
|
| |
There is no use doing these initialization actions in a separate thread,
so let's play safe (races with UI, e.g when setting initial hostname value
as in #902090) and do only waiting for dhcp (for payload initialization)
asynchronously.
|
|
|
|
|
| |
It was trying to mount the nfs path including the file.iso, so strip
that off before passing it to _setupNFS.
|
|
|
|
|
|
| |
Searching for an empty string inside of another string always returns
True. It wasn't properly detecting whether dracut had setup the nfs
repo.
|
|
|
|
|
|
|
| |
Yum properties have side-effects and are inconsistent in their behavior
(eg. one call may go grab something from the net, the next may just return a
list). So we need to be paranoid about wrapping self._yum usage in a
_yum_lock.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
Basically, _setupDevice gets confused when there's a symlink involved in the
path and /proc/mounts does not match what we expect. Thus, just make sure
we resolve symlinks to their real path before feeding them to get_mount_device.
Also, don't print some mount-related log messages if they'd say something like
"None is already mounted".
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Cases I tested:
(spoke is pre-hub network spoke)
Link has (slow) dhcp:
- connected in payload init, connected in spoke
- not connected in payload init, connected in spoke
- fixed by the patch
- not connected in payload init, not connected in spoke
- subcases: - wait for connection in spoke
- leave spoke while connecting
- fixed by the patch
Link does not have dhcp:
- configure static connection and leave spoke
|
| |
|
|
|
|
|
| |
Re-arranges a bit of code and handles dracut mounted hd isos.
Also handles a None type on self._currentIsoFile better.
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
This time it shoves all the logging data into packaging.log
This commit has some hacks to work around yum issues. I've discussed
with upstream and they're going to work on better APIs around logging in
the future.
|
|
|
|
|
|
|
|
|
| |
We could have method of nfs:server:/path/ or nfs:server:/path/foo.iso.
Dracut could have left stuff mounted to /run/install/repo or
/run/install/isodir or both. In any case we should detect what's been
left over and setup yum the right way.
This introduces a couple new constants as well.
|
|
|
|
|
| |
This is not handled by similar code in the software spoke, because that code
is never run for kickstart installs containing a %packages section.
|
|
|
|
|
|
| |
This removes code duplicated in desktop.py, and will short circuit the
routine if the user has indicated that they want a graphical boot by way
of kickstart data.
|
|
|
|
| |
We don't provide you with the chance to do anything anyway.
|
|
|
|
|
|
| |
We are trying to prevent the user from getting to the hub with a space
error, hence why you can only quit. It's still helpful to catch the
errors just in case something sneaks through later.
|
| |
|
|
|
|
|
|
|
| |
We don't want to display an error dialog (and we don't even support displaying
a dialog for this exception). We should instead raise an exception we know
how to handle, and then the status line for the source spoke will be set
correctly.
|
|
|
|
|
| |
We have to potentially unmangle a lang value from say es_ES.UTF-8 down
to just "es" in order to find the group within comps.
|
| |
|
|
|
|
|
|
| |
This will require lorax to actually put the plugin files in the image,
but once they are there we get all the benefit of somebody else's code
managing conditional package installs.
|
| |
|
|
|
|
|
| |
rsync tries to remove xattrs even though -x has been specified. They
don't need to be copied anyway so exclude /dev/ /proc/ and /sys/
|
|
|
|
|
|
|
| |
This is an ugly hack, but it gets nfsiso working. You can switch from
NFSISO to another source, and back to the same NFSISO, just not to a
different NFSISO. That's going to have to be a "known issue" for F18
Beta.
|
|
|
|
|
| |
This adds a thread that uses os.statvfs to monitor the amount data
that has been copied to the destination mountpoints.
|
|
|
|
|
|
|
|
|
|
|
| |
Two fixes here, one set the path to INSTALL_TREE because we haven't
moved anything over to ISO_DIR yet. Two account for the isys mount
count for paths. When we move a mount point the isys count must be
updated so that an attempt later to mount the iso into the INSTALL_TREE
won't be short circuited.
One could create an (_)isys call that does the move and does the
accounting, but that would add more C code, which is an anti-goal.
|
|
|
|
|
| |
It can get it as an attribute off the Payload object, now that setup is
always being called.
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
Otherwise, previously selected packages will have their "selected" attribute
set, so we will not try to mark them for installation again. This leads to
some very strange log output.
|
|
|
|
|
|
|
|
|
|
|
| |
All this does is set skip_broken and yum does all the rest. Note that we
are still doing dep solving a second time, when the progress hub is up, but
adding this setting to the Payload object means skip_broken will be in effect
there too. We still need proper error handling for other errors that might
occur at that point.
This should take care of #865073. I can't think of any cases where you can
get past the first hub with a dep solving error.
|
|
|
|
|
| |
This is a different path than that originally reported, but the
same fundamental cause.
|
|
|
|
|
|
|
|
|
| |
move doesn't work because /run/ is a shared filesystem. So do as
systemd suggests, mark the filesystem as private. We don't know of any
reason in anaconda for the filesystem to be shared anyway.
This hack can go away if/when the kernel ever allows moving mounts
within a shared filesystem.
|
| |
|
|
|
|
|
| |
Otherwise, you'll end up in a situation where packages the previous and
current environment share will not get installed.
|
|
|
|
|
| |
Our forced packages should be marked for installation after all removals
have taken place.
|