future of smart

[Grant McWilliams]

On Thu, Jun 5, 2008 at 5:23 AM, Felix Miata <mrmazda at ij.net> wrote:

> On 2008/06/05 01:37 (GMT-0700) Grant McWilliams apparently typed:
>
> I use Smart cmd line often, because it's what I use for SUSE Factory, which
> I
> keep on about 8 systems. On all, I have restricted their / partitions to
> relatively small sizes. This makes Smart's default behavior to download
> everything before installing anything on 'smart upgrade --yes' a poor
> option,
> as it quite commonly computes that more space on disk is required to
> install
> anything at all. That's BS. There's plenty space if it behaved like urpmi
> did
> by default, automatically finding related packages, downloading,
> installing,
> and deleting in groups of up to 16 or so.
>

Yes, I'd forgot about this. I'm adding it to my list. I've had to work
around a lot of smarts problems with shell scripts. I have a script that
takes a list of all rpms on a base CentOS installation and compares it to a
list of rpms we want on the fully installed system and then goes about using
smart to install the differences. The --stepped module is almost worthless
and would take two days to install the list so I grab about 10 files and do
a normal smart install on them and then because return codes aren't reliable
I have to use rpm to check and see if they installed properly if not I step
through the same 10 files and retry as necessary utnil they're installed and
then go on. If I stepped through them to start with it would take days
because of the cache refreshing which takes longer than actually installing
the package.

So yes I'd like to see smart parse the dependencies and create groups that
could be installed in one go, download the group and install it. What would
be even cooler would be to have it spawn a child process that continues
downloading future groups *if *disk space was adequate. If the group was too
large for the disk space available it could install them one at a time. This
would allow us to get all rpms installed not matter the disk space and as
fast as possible.

My current list


   1. groupinstall like yum
   2. reliable return codes everywhere - having to grep the smartpm output
   for errors is really messy!
   3. fewer cache updates especially when nothing was done.
   4. variables in repo/mirror URL
   5. Retrieve mirrorlist instead of having to have it added manually
   6. Cleaner gui - search button/pane needs to be more intuitive
   7. Cleaner syntax - sometimes it gets really messy.
   8. resolve dependencies and depending on disk space create groups of
   packages that can be installed together
   9. Docs finished
   10. Source packages (like "apt-get source")
   11. Build dependencies (like "apt-get build-dep")
   12. Bootstrapping (like with debootstrap/rpmstrap)

I've added Anders list too.


Grant
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.labix.org/pipermail/smart-labix.org/attachments/20080605/440e7cee/attachment-0003.htm>