[tarantool-patches] Re: [PATCH v2 2/2] Fixed lost format on update and upsert operations.

[Kirill Shcherbatov <kshcherbatov@tarantool.org>]

> Please, describe why we need inlined update here. 'Vlad insisted' is not a reason.

Unlike update and upsert, transform creates a new tuple with runtime format; 
Moreover, the function inlining allowed to make allocations in the region, what is preferable.


> Vova means, that if tuple:transform would use non-default tuple format,
> the tests will pass anyway. Please, add a test on transform, that would not
> pass, if tuple uses non-default format.
> This test will pass even if we would use non-default tuple format
diff --git a/test/engine/update.result b/test/engine/update.result
index b73037e..6ab4436 100644
--- a/test/engine/update.result
+++ b/test/engine/update.result
@@ -728,11 +728,7 @@ aa:update({{'=',2, 666}})
 aa:transform(-1, 1)
 ---
 - [1, 'ssss', '3', '4', '5', '6']
-...
-aa:transform(1, 6)
 ---
-- ['7']
-...
 aa = nil
 ---
 ...
diff --git a/test/engine/update.test.lua b/test/engine/update.test.lua
index e2a2265..3ff7e62 100644
--- a/test/engine/update.test.lua
+++ b/test/engine/update.test.lua
@@ -112,7 +112,6 @@ aa.VAL
 aa:update({{'=',2, 666}})
 -- test transform integrity
 aa:transform(-1, 1)
-aa:transform(1, 6)
 aa = nil
 
 s:upsert({2, 'wwwww'}, {{'=', 2, 'wwwww'}})