I was under the impression repoman can commit changes to tree, not per ebuild. In any case, feel free to edit the script to run repoman for you. I'm sure it will finish over night without your presence at the PC. - so no trouble at all.
repoman can do commits on a per-ebuild basis (rather, per-package), or it can be done from a higher level for multiple package commits at the same time. However, repoman runs an extremely detailed set of checks on packages, and this can take time. Dan Armak reported that a repoman run on the entire KDE category takes about 9 minutes, and this is likely on a rather modern x86 box. Doing the same thing on a slower mips box (or other non-x86 box, like sparc) will take considerably longer, and opens the door for someone else to make a quick change to a KDE thing while our repoman is running, thereby botching the entire commit (there are other ways around this, though).
I still have no idea how its any different to testing kdemultimedia instead of kdemultimedia-meta in big endian and little endian installs over three different userland ABIs. As far as I can tell, there isnt a single difference.
I reccommend you go and purchase the book "See MIPS Run" by Dominic Sweetman off of Amazon, and read up on the MIPS architecture. There are also MIPS-related documents on ftp.linux-mips.org, especially of note is the N32 handbook. Once you've understood a little about other architectures, then you'll be in a better position to understand the concept of multiple ABIs on a given system.
On MIPS, our basic, run-in-the-mill ABI is o32, basically the kind of userland you'll find in IRIX 5.x. It's pure 32bit, and doesn't understand anything about 64bit. It's also the most stable, and thus, most widely used under Linux. There's also n32, commonly found in IRIX 6.x. It's a hybrid -- it's got the ability to do many things with 64bit performance w/i incurring your usual 64bit penalties. It's very experimental in Linux atm, still requiring some more hackery to glibc, gcc, and probably the kernel.
There's also n64, which is the MIPS-equivalent of pure 64bit (no 32bit stuff). While this one isn't in very wide use under Linux (and probably only certain apps in IRIX use it), it's still an ABI that will quite likely have a future use, and will thusly need testing.
Now, with that knowledge somewhat provided to you (I highly suggest you look up the rest on your own), re-consider your statement about how you see no difference in testing KDE over those ABIs (and even different endians).
--Kumba
--
"Such is oft the course of deeds that move the wheels of the world: small hands do them because they must, while the eyes of the great are elsewhere." --Elrond
-- gentoo-dev@gentoo.org mailing list
