Re: [StGit PATCH] Convert "pop" to the lib infrastructure

On 2009-04-02 17:20:45 +0100, Catalin Marinas wrote:

2009/4/1 Karl Hasselström [off-list ref]:

quoted

On 2009-03-31 12:30:27 +0100, Catalin Marinas wrote:

quoted

@@ -47,7 +47,7 @@ test_expect_success 'Try to reset with --hard' '

     stg reset --hard master.stgit^~1 &&
     stg status a > actual.txt &&
     test_cmp expected.txt actual.txt &&
-    test "$(echo $(stg series))" = "> p1 - p3 - p2"
+    test "$(echo $(stg series))" = "> p1 - p2 - p3"
 '

Hmm, why this change in behavior? Something that should be noted
in the commit message?

quoted

@@ -47,7 +47,7 @@ test_expect_success 'Try to undo with --hard' '

     stg undo --hard &&
     stg status a > actual.txt &&
     test_cmp expected.txt actual.txt &&
-    test "$(echo $(stg series))" = "> p1 - p3 - p2"
+    test "$(echo $(stg series))" = "> p1 - p2 - p3"
 '

And I guess this is the same.

I think we now get a slightly different behaviour because of how the
transactions are generated with the new infrastructure. In the above
case, you have "pop p2 p3" and "push p3", the latter failing. The
"pop p2 p3" command results in the stack being "> p1 - p2 - p3"
while "push p3" performs a single step for pushing and reordering.
The old push caused a reorder followed by a push.

Ah, OK. Hmm, I guess either behavior has its pros and cons. (Though I
guess the new behavior -- not changing the order when the push failed
-- might be slightly more intuitive.)

Add that explanation to the commit message, and I'll award you a

Acked-by: Karl Hasselström <redacted>

-- 
Karl Hasselström, kha@treskal.com
      www.treskal.com/kalle