On 15 Sep 2026, at 23:12, Devin Teske <[email protected]> wrote:
> 
> 
> 
>> On Sep 15, 2026, at 3:08 PM, Jessica Clarke <[email protected]> wrote:
>> 
>> On 15 Sep 2026, at 22:55, Devin Teske <[email protected]> wrote:
>>> 
>>> Splitting leaves landmines for bisection.
>>> 
>>> The bundled pieces have to land together so main is not left half-broken 
>>> for bisect.
>> 
>> From what the commit message says I find it hard to believe that it
>> would not be possible to split it into multiple commits without
>> introducing new, temporary regressions. Is that what you are trying to
>> say here?
>> 
> 
> Yes. That is what I meant by landmines for bisect.
> 
> Some of the bullets could have been separate commits on a branch if
> ordered carefully. I landed one Fixes commit for eaeb5f29bc6f so main
> is not left with a half-repaired load path between tips.

Half-repaired is fine, you can have multiple Fixes: commits. Having
those "separate commits on a branch if ordered carefully" be the actual
commits landed in main is generally best practice, both within the
project and generally within software engineering. So if that’s
possible, does that not contradict your statement that splitting would
leave such "landmines", without any qualification to make it anything
but an absolute statement?

Jessica


Reply via email to