Still unable to use Smart with Mandriva

[Anders F Björklund]

Jeff Johnson wrote:

>> Is this the real smart 1.4, or the "mandriva" branch ?
> 
> How serious is the divergence?

It's two different code trees by now:

https://code.launchpad.net/~smartpm/smart/trunk

https://code.launchpad.net/~proyvind/smart/mandriva

 29 files changed, 1237 insertions(+), 160 deletions(-)

>> Previously Mandriva have been using their own fork,
>> which had some issues like the one you are describing...
>> I merged the applicable code, and added some workarounds
>> for where it had a LZMA / XZ file as a .xml, and so on.


> And I will be undertaking to try to smooth out the extra '-' added
> to package file names as a result of Distepoch: and Disttag:.
> 
> There are only two resolutions:
> 	Keep the '-' in %___NVRA.
> 	Remove the '-' in %___NVRA
> 
> I'm inclined to removal (and I'm pretty sure that I can write code
> to have its BOTH ways as while transitioning).
> 
> There's pending refactoring in RPM that is log jammed on the keep/remove decision.
> A two-tailed coin will be flipped if I must pretend at "Have it your own way!" choice.

I was adding the Distepoch and Disttag to the version,
so that it matched the filenames (i.e. the double dash)

e.g. rpm-5.3.12-0.20110712.2-mdv2012.0 at i586

The result differs from which rpm version that you use,
that is whether it knows about Distepoch/Disttag or not.

e.g. rpm-5.3.12-0.20110712.2 at i586 (rpm.org)


And I don't want any Distepochs in the Depends, so the
version comparison disregards those and only uses "EVR":

rpm = 1:5.3.12-0.20110712.2

This meant that all the "strcmp" (or Python equivalent)
had to be replaced with "vercmp" for doing comparisons.


Seems like some "synthesis" fixes are still needed,
but otherwise the "dist-aware" code is living at:

https://code.launchpad.net/~afb/smart/dist

--anders