Newbie question with Smart

[Anders F Björklund]

Sriram Narayanan wrote:

> Hi Anders, thanks for the tips. I got this working now.

Glad to hear... (saw your post)

> I had to place "i86pc-pc-Solaris" in /etc/rpm/platform (I'm using a 32
> bit build for the time being).

The OS should probably be lowercase, and the vendor "something" ?

Typical values would be i686-redhat-linux or x86_64-apple-darwin,
so I would have expected a "i86pc-belenix-solaris" (or somesuch).
Whether it is "solaris" or "opensolaris", I don't really know...
(guess it's a similar discussion to "linux" vs "linux-gnu" etc)

Arch is up to you, I know that OpenPKG is using "ix86" for theirs.

>> 
>>> I've added a channel, and am able to run an update, but I'm unable to
>>> install the package. What should I investigate ?
>> 
>> You should look into that "rpm.archscore" returns something
>> for the package. Otherwise it won't be added to the channel:
>> 
>> $ python
>>>>> import rpm
>>>>> rpm.archscore("i86pc")
>> 0
>> 
> 
> Yes, I get this very output, but I don't know what it means. I need to
> study rpm internals and read the docs better!

A score of 0 means incompatible, otherwise lower is "better".

The score is used in two places, the first is to reject all
packages with a score of zero to cut down on the number (?).

	if rpm.archscore(arch) == 0:
		continue

The second is for sorting packages with the same version, so
that the "best" arch package for the platform is selected.

	if rc == 0:
		rc = -cmp(archscore(selfarch), archscore(otherarch))

You will find the actual definition of the score within RPM.

>> If you are running rpm5, you need to set up /etc/rpm/platform
>> since that is what it is using to determine platform "score":
>> 
>> rpm --eval "%{_arch}-%{_vendor}-%{_os}%{?_gnu}"
>> # See also https://bugs.launchpad.net/smart/+bug/587448
>> 
> 
> As noted above:
> 
> I had to place "i86pc-pc-Solaris" in /etc/rpm/platform (I'm using a 32
> bit build for the time being).

The "i86pc" is pretty non-standard, but already in place for
debs (it links to "i386" deb packages, ditto sunos5 -> solaris)

The main issue in that bug was that the %_gnu value is ignored,
which only matters on ARM after the introduction of -gnueabi...

--anders