On Thu, Sep 27, 2007 at 02:46:09PM -0700, ron minnich wrote:
On 9/27/07, Robert Millan email@example.com wrote:
- include: someone might break board A in a commit that was only tested on board B.
I can only say it's been a huge problem in the past. I've had cases where I tried to rebuild linuxbios for an older board, only to find that "improvements" in "common" code rendered the board unbootable. this is a particular problem for supermicro motherboards, which change in trivial ways all the time. An irq table that is common to two boards may not be a few months later. If the newer board is changed, the older board is broken. This really happened to me.
Sure, there will be such cases, but that's a bug then anyway. If someone breaks the code that's either (a) a bug, or (b) a sign that the "common" code wasn't really common in the first place and should be moved respectively.
This should not affect _users_ all too much (there's always svn, you can always checkout older/working versions, see also http://linuxbios.org/Confirmed_working_svn_revisions, which we could expand maybe). We also have abuild, and the hardware test suite which (I hope) will have a bunch of machines integrated sooner or later.
Upon the occurence of such an issue the _developers_ just Fix It (tm) and that's it. Not a big deal, IMHO.
Note: I _do_ understand your reasoning, and yes, this type of bugs may occur, but common code is still the better route IMO, and svn will save our ass until such bugs are fixed...