Will; On Mon, Aug 20, 2007 at 05:26:48PM -0400, William Cohen wrote: > Stephane Eranian wrote: > > I am have been able to the current git clone built and perfmon2 is working > on x86_64 machine. I am not that versed in git, but I am starting to take a
Good to hear, it works for you. > look over the git tree via the web page. Was there a reason that the > perfmon patch was done as one large merge rather than a set of smaller > patches? Is there some way to break out the smaller patch sets, e.g. > arch-independent patch and the various arch dependent patches? I hesitated for a long time about this particular issue of how to check in the patches. As you know, the ususal release contains a series of patches: common vs. per arch, new vs. modified files. The key here is that with the current breakdown, each separate diff file is not independent of the other. In other words, you cannot not apply just one and expect to compile the kernel without errors. There are just too many dependencies between the various layers. For GIT, I talked to Andrew about this. Initially I thought of committing each of the diff files separately. But that would not work because you would not be able to compile a kernel without having ALL applied. Then, there was the issue of backtracking patches when you find a regression (git bisect). If you cannot compile the kernel when you undo one perfmon commit, then you cannot easily isolate the cause of a regression. The alternative approach would be to rework the patch structure completely to try and isolate independent pieces. That would have been very difficult and time-consuming. Andrew said that we might as well try to push it in all at once (to make bisect work). -- -Stephane _______________________________________________ perfmon mailing list [email protected] http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/
