On Fri, Mar 4, 2011 at 4:34 PM, Gabe Black <[email protected]> wrote:

>
> > In general, we need to be much more cautious about updates that break
> > existing checkpoints.
>
> [...] Caution was not a factor.
>

Yea, that's my point ;-).


> As I said last time this came up, I'll won't take breaking checkpoints
> lightly, but my primary goal is improving the code as it is today, not
> compatibility. If I need to fix a bug or have a change that
> significantly enhances functionality, I'm going choose to do that over
> keeping old checkpoints working. I understand the changes can be
> painful, but it would be a mistake to trap ourselves like that.
>

I think you're misunderstanding me... I don't mean cautious as in "don't fix
bugs if you can't do it without breaking existing checkpoints".  I'm
thinking of things like:

- adding sensible defaults (when possible) to new unserialize statements, so
that (when possible) things work automatically instead of having
hypothetical post-facto email discussions about it ;-)

- when necessary, adding warnings to commit messages that say "THIS WILL
BREAK EXISTING CHECKPOINTS" so people don't find out only after they've
incorporated a changeset that they have just committed themselves to a bunch
of extra work

- working on that previously discussed python script that could automate
checkpoint updates when simple defaults are insufficient

I think a big factor in making this happen would be having a regression test
that explicitly loads from a static checkpoint, so that in the process of
running regressions it would become obvious that you've broken that
capability.

Steve
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev

Reply via email to