Redo the rpmdiff behavior?

[scarabeusiv]

With difflib from python core we could generate unified diff look alike for the rpm changes instead of printing added/removed....

Do we happen to have some set requirements for what rpmdiff should actually do? Also some of the file attributes diffing is dependant on the underlying FS so it might be good idea to create tmpfs where we would do the files comparison...

[scarabeusiv]

@dirkmueller @Conan-Kudo What would be your opinions on the above? Also note that any changes in the cli interface will have to come when I finish with reducing the pkg.py blobie.

[Conan-Kudo]

I like the idea of a unified diff-style view, as it's something people can easily comprehend on changes.

Could we also do that for RPM payload information (metadata, deps, etc.)?

[Conan-Kudo]

Also, I vaguely recall that Koji has a vendored copy of rpmdiff. I don't see why Koji shouldn't just use ours, now that it's usable as a module.

@mikem23, @tkopecek, what are your requirements from rpmdiff for Koji?

[tkopecek]

Very briefly, we do a comparison of noarch rpms generated on each architecture and looking for significant differences (rpmdiff -S5TN). Further we're caching hash of this data, to compute it only once. This is probably the most significant difference. Otherwise, it is quite simple rpmdiff-like code (https://pagure.io/koji/blob/master/f/koji/rpmdiff.py)