On 3/16/2021 1:25 PM, Elijah Newren wrote:

On Tue, Mar 16, 2021 at 10:01 AM Derrick Stolee [off-list ref] wrote:

quoted

On 3/16/2021 12:05 AM, Elijah Newren via GitGitGadget wrote:

quoted

This tiny series depends on ort-perf-batch-10[1].

If the ort-remainder topic[2] is merged with this series, then the result is
a version of merge-ort ready for general usage. Users can select it by (a)
passing -sort to either git merge or git rebase, or (b) by setting
pull.twohead=ort [3], or (c) by setting GIT_TEST_MERGE_ALGORITHM=ort.

Does the other topic add GIT_TEST_MERGE_ALGORITHM=ort to the CI builds?

Specifically, the Linux builds have a second run with some optional
GIT_TEST_* environment variables. This seems like a nice addition.
 Other than that extra request, this series was easy to review. LGTM.

The other topic left tests in t6423 failing.  This topic leaves the
tests in t6409 and t6418 failing; it's only the merge of the two that
has all passing tests.

Combining the series might be the right call, or say this one depends
on that one.
 

I guess since the ort-remainder topic still hasn't been picked up, I
could just combine it with this series (basing on ort-perf-batch-10).
Then I could either add a patch at the end of the series that runs
tests under GIT_TEST_MERGE_ALGORITHM=ort, or which changes the default
merge backend to ort. 

Perhaps change the backend to "ort" only when "features.experimental"
is enabled, at least for one full release? I'll do my part to do some
testing with our repos in that time, too.

Thanks,
-Stolee