On Sat, Mar 6, 2010 at 5:39 PM, Paul Fisher <r01.gjk@xxxxxxxxx> wrote: > This is probably a stupid question, but really, is it still possible? How could it not be? > Yes, the rate of change is very high, Andrew says that it is scary, > but those changes > are made by pro developers, and they don't do silly mistakes. Of course they do. The real secret to the linux kernel quality is the open process and the concept of maintainers. I've been watching the ext4 portion of the tree as an example recently. There are lots of contributors even to this one relatively small portion of the kernel tree. The official maintainer is Theodore T'so, but there are several other very knowledgable people he clearly leans on. Everyone including Ted posts their patches to the public ext4 list for review by all interested parties prior to there being added to the official ext4 test tree. I think only Ted adds patches to the ext4 test tree, but he may share that responsibility with others. After they have sat there a while and Ted is satisfied he pushes them to Linus for inclusion in mainline. There is nothing that stops you from reviewing the ext4 code and submitting patches via the ext4 mailing list. If no has issues and Ted accepts the patch, it should eventually make it into mainline. Same for each of the subsystems. Be aware it is not unusual for major functional patch series to bounce around for a year or two or even more before everyone is happy with them and they eventually get in. The drbd patch that went into 2.6.33 is an example of that. They first submitted their code a couple years ago I think. It was a major patch, so it had tons of feedback and required lots of tweaking before it was eventually accepted. And of course it was not a sure thing it would ever be accepted. > 75% of all > contributors are paid to do this job: all those professionals work on it full > time, they probably have a whole testing teams behind there back, or at least a > small network of server-class machines at their basement. Just because their paid, doesn't mean its full time. Certainly some are. Many are not. Most do NOT have a network of server-class machines I don't believe. There are labs like OSDL that make their equipment available via remote connection for kernel developer use. I don't know how you get access, but the whole point is to provide free resources to kernel hackers. > The kernel is > historically supposed to be bug free, and it seems to be true (except for > drivers for some obscure hw). Many eyes and lots of testing by the community, but even then errors get in. > Andrew keeps saying that number one project for all kernel newbies should is to > make sure that kernel runs on their hw (because so much different permutations > is possible), but the problem is that most of us currently have a very typical, > standard hw and there is no way to find a regression with such a setup. > > Greg encourages to start with staging tree cleanups, but it is really good only > for understanding a submission process. Once you get it, you want to some > _real_ work and since staging tree contains drivers, you almost always need a hw > to test your changes. Of course some people do post untested patches, but I just > can't do that. > > So, again, is it still possible to start somewhere and go on? I mean not doing > some mechanical job, but something that is useful to the community (of course > assuming that one is not a complete newbie, knows C, a bit of assembly, have > read LDD and UTLK3)? If this is really a stupid question, please just say it and > nothing more, I don't want any flames here. One can say that this mail was a > waste of time, and that I should have spent this time actually working on the > kernel, but to me this is an important thing to know. I encourage you to check out ohsm as one example. We are trying to develop a new feature/subsystem. For now it is focused on ext4, but we could add xfs support also. Maybe btrfs support even makes sense. See http://ohsm.sourceforge.net We still have lots of work to do and have not yet submitted anything for official review. And when we do who knows if it will be accepted or not. > --paul Greg -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
