On Mon, Sep 24, 2012 at 12:23:10PM -0500, Joe Hershberger wrote:
Hi Tom,
On Mon, Sep 24, 2012 at 12:18 PM, Tom Rini trini@ti.com wrote:
On Mon, Sep 24, 2012 at 07:09:35PM +0200, Marek Vasut wrote:
Dear Tom Rini,
On Mon, Sep 24, 2012 at 06:53:19PM +0200, Marek Vasut wrote:
Dear Joe Hershberger,
[...]
While you're touching all these files, why not replace the __u_boot_cmd* implementation with this as well. There's no need to leave a special case in there as well is there?
Not in the first stab, it can be indeed done later, but now I'd like to avoid breakage. Besides, replacing u_boot_cmd would break bisectability, I want to preserve it and rather apply more patches slowly than less patches recklessly.
Agreed. We need to work hard to make sure we do lots of incremental steps here to make sure breakage can be bisected.
+1 (actually +inf.)
Making use of the 'exec' step in git rebase is very helpful for making sure this is the case too, btw.
exec step ?
Yes, in 'git rebase -i' you can insert exec lines after each commit such as: pick 12345 Commit 1 exec script-that-runs-MAKEALL.sh pick 6789a Commit 2 exec script-that-runs-MAKEALL.sh
I use my MAKEALL wrapper (http://pastebin.com/fNhG4iCd but I've updated a bit more since I uploaded that) and --log `git rev-parse --short HEAD` so I can see what's gone on for every step in a series.
That's interesting. I've been using git-test-sequence for a similar purpose. It doesn't require the exec steps to be added manually, though.
+1 to git test-squence as well. I think the reason I didn't use that after a while was that MAKEALL didn't non-zero exit on warnings. Of course since I've moved on to logging all output from each step to verfiy by hand, test-sequence would also work great here.