Red Hat Bugzilla – Bug 487584
i386 package installation via SDC fails on x86_64
Last modified: 2009-08-13 16:39:59 EDT
Description of problem:
If a user attempts to install an i386 package on an x86_64 client via the SDC, the x86_64 equivalent will instead be installed. For more information, visit:
Version-Release number of selected component (if applicable):
Steps to Reproduce (copied from BZ #456450):
1. select Systems
2. select an x86_64 system
3. select Software -> Packages -> Install
4. select a Package that exists for both x86_64 and i386 (e.g. libjpeg)
5. schedule package install
6. run rhn_check -vvv on client at appropriate time to install the package
Actual results (copied from BZ #456450):
On 3, observe that the package list does not indicate package architecture.
After 6, observe that the x86_64 version of the package has been installed.
Expected results (copied from BZ #456450):
On 3, user should be able to select any package that is available to the system;
therefore, if both i386 and x86_64 is available, they should see both listed.
After 6, the package selected (e.g. i386) should be installed
For step 3, are you indicating that there is no column for "Architecture" or that the column exists, but it's content is empty? If possible, please provide a screenshot or send me a URL to the server so that I can log in an look at the behavior you are seeing.
Also, what is the client version being used?
(Note: must be RHEL4.8 or 5.4 in order for the multi-arch enhancements to be fully functional)
Whoops... I think that you caught a bug in my bug. Disregard everything related to Step 3 in the Actual/Expected results, as the Architecture column shows up and is appropriately populated by each package's respective architecture.
I'm using RHEL 5.3 on my x86_64 client... to the best of my knowledge, 5.4 is still in its development phase, though I can certainly wait for it to move to QA to test on RHEL 5 should 5.3 be inadequate.
That is good regarding step 4.
If possible, you might run the same test, run rhn_check -vvvvv from the client and capture the output from rhn_check. This should show us what information is being communicated to the client. If Satellite does not communicate the arch of the package (e.g. i386) in this case, then it may be a problem on the UI side; otherwise, it could be a problem in the client in which case we can check with Pradeep to see if this is a case that should work in RHEL 5.3. (There are some fixes for multiarch that went in in RHEL 5.3 and others that went in to 5.4.)
Upon further review, this test succeeds on RHEL 5.3, and I've attached the output from rhn_check and rpm -q to verify this success.
Per our chat, please let me know if the behavior has been confirmed with a RHEL 4.8 client.
Correct behavior confirmed on a RHEL4.8 client too; moving to CLOSED, NOTABUG.