Since rpm5.org seems headed for a rpm-5.2.0 release in
early May, its's time to sort out API/ABI issues.
I've been asked this week (privately)
What is the defining difference between rpm-5.1.8 and rpm-5.2.0?
The answer is both
There's no essential difference between rpm-5.1.8 and rpm-5.2.0
at a functionality level (as of last week before embedding tcl/perl/
python/ruby)
as well as
The single largest difference between rpm-5.1.8 and rpm-5.2.0
is in the API/ABI.
The difference started roughly last August in this thread
http://rpm5.org/community/rpm-devel/2859.html
I was asked (by arekm from PLD) to commit to an stable API/ABI
@rpm5.org/
In order to do that, I switched rpm-5.2.0 from a "maximal" approach
used in all releases of rpm back to rpm-4.1 (i.e. you get to see
everything
that rpm uses) to a "minimal" approach to API/ABI for perfectly
obvious reasons:
You don't need some #include <foo.h> if its not exported and its an
internal interface.
Meanwhile, the single largest difference (if diff'ing code) between
rpm-5.1.8 and rpm-5.2.8 is that rpm-5.1.8 uses <stdint.h> integer
names like "unit32_t" while rpm-5.2.0 uses internal typedef's like
"rpmuint32_t". The reason for the diffence is described at the URI
I've given.
The issue is largely cosmetic; indeed 32 bit ints have 32 bits
no matter what RPM decides to call them.
However, there is most definitely a large and significant difference
between "minimal" (as in rpm-5.2.0) and "maximal" (as in previous rpm
releases).
I personally don't think that rpm has any API/ABI worth exporting except
through bindings (which is a whole different matter) while there are
multiple rpm fork's active. But YMMV, everyone's does.
Since rpm-5.2.0 is heade for release in early May, its important
to identify what is desired as soon as possible.
Any reasonable request will be accomodated.
hth
73 de Jeff
______________________________________________________________________
RPM Package Manager http://rpm5.org
Developer Communication List rpm-devel@rpm5.org
