RPM Development Priorities over the next 5 months or so
These items came out of ongoing discussions with rpm and yum developers. The sign
denotes whether a feature is finished in the current git RPM version.
RPM
Internal design improvements
implement dynamic macro expansion, current implementation passes fixed sized buffers to expand macros
fix stack buffer overflows wherever they may be found
alloca() is no more used in the whole RPM code
improve memory use/performance during transaction
File finger printing (FlorianFesti)
Not important but hopefully low-hanging:
RPMTRANS_NOCHANGELOG or RPMTRANS_CHANGELOG_LIMIT
OR changelog as (additional) regular %doc file?
RPMTRANS_NOFSSYNC
Features
LZMA support
only the new LZMA format (with magic at the beginning) is supported
allow noarch subpackages
%{isa} functionality for manual and automatic deps
dependency/provide filters in build process
spec-global provide+require regex for filtering
Add "hdr" manip. code for key/val pairs that yum wants
from_repo
note
installed_for (groupinstall+group/dep+dep/explicit - enough?)
Correct file finger printing (FlorianFesti)
Compatibility
drop internal lua
drop internal Berkeley DB
remove requirement of Group: - it must stay, however
ignore BuildRoot in spec files, rpm should know best itself
default fuzz factor for patches set to 0
provide a mechanism to quickly see if the rpmdb has been modified (for caching rpmdb info w/o
having to constantly recheck the indexes)
RPM-Python-NG
Ability to create/write an rpm signature and package hdr objects and serialize them
Ability to access hdr manip. code for k-v pairs from above
Transaction Callback code - from rpmtrans.py and rpmUtils in yum
code to check rpm package signatures w/o using rpm itself and using a keyring of keys
a lot of the standard stuff that's in rpmUtils.miscutils
Related
See Yum Development Priorities:
http://wiki.linux.duke.edu/YumDevPriorities