Re: YUM install [package.rpm] fails on Solaris, reports success

Hi, 
 
By "last", I meant the processing of %post in any given package. Currently we're just testing with a small number of packages, and %post always fails in each (some problem with the creation/manipulation of /var/local/tmp/rpm-tmp.number). We're not using %pre yet but I suspect this may also fail as a result of the same issue. 
 
I understand that "success" is a subjective term, and that scriptlet failure may not be considered fatal. In this case, because we are dealing with packages across 3[+] platforms, some of the initial packages we are working on are [noarch] and mostly logic/configuration file manipulation, so if the script fails, nothing substantial remains other than the entry in the rpm db. I'm guessing "success/failure" is probably something that is reported by rpm-lib rather than declared by yum at any rate. I think for us, though, this issue is probably pretty significant in terms of the prognosis for the success of the project.
 
-j

--- On Thu, 11/12/09, Seth Vidal <[email protected]> wrote:


From: Seth Vidal <[email protected]>
Subject: Re: [Yum] YUM install [package.rpm] fails on Solaris, reports success
To: "Yellowdog Updater, Modified" <[email protected]>
Date: Thursday, November 12, 2009, 6:01 PM




On Thu, 12 Nov 2009, Joshua Burns wrote:

> Hi guys,
> 
> Am working on a major project at a very large company to standardize on YUM/RPM throughout the UNIX/Linux environment
> globally. Linux works out of the box, of course, so I am working on getting Solaris working next. AIX will be last.
> 
> Have compiled RPM and YUM on Solaris 10 (sparc), and have tested RPM fairly thoroughly - it appears to work fine, and
> package installation works well.
> 
> Yum also seems to be complete, and I can get repo listings and the like, but when I attempt a yum install <pkg> , the
> last (or one of the last) steps fails, when the %post script is supposed to be copied to /var/local/tmp/rpm-tmp.number
> and executed. This fails, evidently because the fd manipulation is handled differently (when using yum versus rpm). The
> yum utility reports success, and in fact the package appears in the RPM database, though the %post processing does not
> take place. Installing via RPM, however, succeeds.
> 
> A truss(1) shows that the sequence of function calls between the two (rpm install versus yum install) varies in terms of
> fd function calls for rpm-tmp.xyz and their sequence.
> 
> Not being intimately familiar with yum/rpm architectural details, I assume that this is because yum is using
> _rpmmodule.so or another rpm library or otherwise handles rpm installation using a different process than using the tool
> directly? BTW, I have scrupulously verified things like directory permissions and the like - I do not believe this is the
> issue.
> 
> Details:
> 
> OS: Sun Solaris 10, on sparc
> Python version 2.5.4
> RPM version 4.4.2.3
> Yum version 3.2.25
> (I can give exact versions for all of yum's dependent packages if needed).
> 
> Exact error message:
> /var/local/tmp/rpm-tmp.30838: /var/local/tmp/rpm-tmp.30838: cannot open
> 
> I can also provide truss information, yum -d 9 -e 9 --verbose does not really provide detail during the %post phase of
> the install. I also have truss information if that is useful.
> 
> Thanks for any help you can give.
> 

All of the transaction is handled internally to rpm-lib through the rpm python module. Yum just makes the calls. Is it only one pkg that is failing or is it always the last pkg?

Secondly, %post scriptlet failures are not fatal to the transaction nor to the pkg being installed.

-sv

_______________________________________________
Yum mailing list
[email protected]
http://lists.baseurl.org/mailman/listinfo/yum


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com

_______________________________________________
Yum mailing list
[email protected]
http://lists.baseurl.org/mailman/listinfo/yum