On 5/1/11 12:35 AM, Karen Tung wrote:
Hi Drew,

The changes for 7040795 look good.

Thanks for looking!


Looking at the changes, and looking at similar code
in the CleanupCPIOInstall
checkpoint in the new CUD architecture, I wonder whether
we need to update line 311-312 of usr/src/lib/install_ict/cleanup_cpio_install.py
so that uuid for all post install publishers are reset instead of
just the highest rank publisher.  I know this does not have to be resolved
right now, perhaps we should track it with a bug so we don't forget.

Actually, what happens in cleanup_cpio_install.py is exactly what we want to happen. The old ICT code doesn't use the API, however, so we're forced to use the CLI. If it's determined that we need to reset the UUID for all publishers, we can certainly go that route. For now, as you've said, I'm going to wait.


Regarding the changes to 7040909, it looks OK in terms
of being able to fix the issues at hand.  Please update the comment
for the function in line 320-324 so it better reflects what's happening.

Will do.  Thanks for looking!


Now that there's no notion of a preferred publisher, I wonder whether
the code for setting of post install publisher can be simplified.

Previously, we are required to have a preferred publisher.  So, the
code for setting the post install publisher is written to first replace the preferred publisher with the preferred publisher for post-install. Then, all additional publishers are deleted and finally add all the additional post install publishers.

Now that we don't have a preferred publisher, do you know whether
we are required to have at least 1 publisher?  If we are not required to
have at least 1 publisher, we can make the code for setting post install
publishers much easier to read and understand by first removing all existing
publishers, then, adding in the new list.  Again, I understand
that you must get this fix into the gate asap.  So, perhaps you can also
track this investigation in a bug, so it can be addressed in the near future.

Actually, there's a few places in ips.py that refers to the notion of a 'preferred' publisher, outside of what I fixed. I'll file a bug for the revamp of ips.py to handle the cleanup and streamlining of the post-install publisher settings.


Thanks,

Thank you for looking, Karen.  Especially late on a Saturday night.

-Drew

_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss

Reply via email to