[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] Switch to GIT. Why?

From: M. Warner Losh
Subject: Re: [Qemu-devel] Switch to GIT. Why?
Date: Mon, 27 Apr 2009 08:42:22 -0600 (MDT)

In message: <address@hidden>
            address@hidden (Lennart Sorensen) writes:
: On Mon, Apr 27, 2009 at 07:54:32AM -0600, M. Warner Losh wrote:
: > Yes.  And that's why you can only check out entire trees, and never
: > subtrees, with git and hg, but you can with svn.  git and hg force a
: > small view of the system to be its own repo.  So something the size of
: > X11 or FreeBSD could never use git or hg without breaking the
: > integrated tree model.  FreeBSD used svn to not break it because we
: > like the development model we have (and think Linus' talk at google
: > was very arrogant to suggest it was broken), while the X11 folks broke
: > their tree up into lots of modules, and got out of the making sure it
: > all worked together business (which makes integrating X11 into systems
: > harder now, alas).
: It is certainly a tradeoff.  Of course with svn, when you do a checkout,
: you get one version of the code.  If you want to search te history,
: you have to talk to the server.  With git a "checkout" (clone) gets you
: everything, history and all.  Makes searches and switching revisions
: much faster, but makes the initial copy take much longer.
: What X11 did also made it much easier to fix bugs, since you can now
: work on the smaller pieces and release them much easier than you can
: release the whole thing.  As for making sure it all works together,
: I am not convinced it made it harder.  Seems development and such has
: gotten much faster and better as a result.  It is no longer completely
: overwhelming to work with.

Heh, it never seemed overwhelming to me when I worked on it...

: > For qemu, none of these issues are likely to be an issue, but it is
: > something to keep in mind...
: True.  Of course as long as you have one tree per application, even
: freebsd should be able to work with something like git.  Having one
: giant tree for everything sounds nuts to me.  No idea how they do it
: (I can't stand the userland of BSD in general, so I don't work with it.
: Good applications, good kernel, lousy setup. :)  Debian kfreebsd may
: change that of course.)

One big tree works very well.  One command builds the whole thing so
updates are easy, but you can still go to a specific directory and
just build that if you find a bug in ls.  Things are stable enough
that it is rare you'd want to just update one thing for a critical bug

It actually works very well and eliminates the version dependency
problems you run into if you are using/building linux packages of
whatever flavor...


reply via email to

[Prev in Thread] Current Thread [Next in Thread]