OK, mystery solved.  Tom noticed a bug in the code section referred to 
below.  Hence we didn't see the "prior_version=" keyword in our log 
files.  He's filing a pkg(5) bug.  Once that's fixed, our adopters will 
be happy to be able to use this keyword to differentiate between new 
installs and updates.

Chris

Shawn Walker wrote:
> Christopher Kampmeier wrote:
>> Adopters of UC2 and multi-platform pkg(5) would like to report on new 
>> package installs vs updates of already installed packages, but it's 
>> not clear to us how to do so.  Currently,  a "pkg install pkg-java" 
>> operation that effects an update of the already installed pkg-java 
>> package results in the following access log record:
>>
>> 129.150.36.226 - - [27/Jan/2009:10:02:17 -0800] "HEAD 
>> /dev/latest/manifest/0/[email protected]%2c0-20.1684%3a20090127t041533z 
>> HTTP/1.1" 200 - "-" "pkg/7570d3749c6f (darwin i386; 9.6.0 Darwin 
>> Kernel Version 9.6.0: Mon Nov 24 17:37:00 PST 2008; 
>> root:xnu-1228.9.59~1/RELEASE_I386; user; pkg)" "-" 
>> "(operation=install;reason=process)"
>>
>> Note that the "operation=install" setting is the same as what is used 
>> when a new package installation occurs.
>>
>> Is there a straightforward means based on log inspection to 
>> distinguish between new package installs and updates to existing 
>> packages?
>>
>> If not, is there support for an RFE to indicate explicitly in the 
>> HEAD requests whether an operation is a new install vs an update?
>
> The trick is that a keyword will appear in the intent header called 
> "prior_version" if a previous version was already installed (look at 
> lines 237-252 of modules/client/retrieve.py in the gate tip).
>
> As for operation=install being there for an update, that will only 
> happen if you're not using image-update to update a package.  Install 
> is indeed the operation.
>
> Cheers,
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to