Improve test coverage for releaseRange rollbacks. #37
Conversation
Codecov Report
@@ Coverage Diff @@
## master #37 +/- ##
=========================================
Coverage ? 85.14%
=========================================
Files ? 9
Lines ? 1851
Branches ? 0
=========================================
Hits ? 1576
Misses ? 163
Partials ? 112Continue to review full report at Codecov.
|
tx_test.go
Outdated
| @@ -602,6 +602,108 @@ func TestTx_CopyFile_Error_Normal(t *testing.T) { | |||
| } | |||
| } | |||
|
|
|||
| // Ensure that a update that releases pages in ranges between read transactions | |||
There was a problem hiding this comment.
To follow go-style function comment, do TestTx_releaseRange ensures that...
There was a problem hiding this comment.
Also I think the test comment should say what we mean by "properly". In particular, this is a correctness test that verifies we don't "overfree". (Unless I read the code wrong.)
| } | ||
| } | ||
|
|
||
| put("k1", "v1") |
There was a problem hiding this comment.
the test sequence is deterministic. we could randomize the test sequence to increase our confidence. but i think it is OK for now. better than nothing.
There was a problem hiding this comment.
Yeah, definitely better than nothing. Could leave behind a TODO comment about that.
There was a problem hiding this comment.
Good questions. Yes, the goal of this test is fairly modest-- Guarantee that we have coverage of a releaseRange being applied to the txid ranges between long running transactions and that we check that the read transactions can read values from the txids they are pinned to after the release.
I'll flesh out the documentation of this test to make this more clear.
I'm working on a separate quick (randomized) test that might be better used for soak testing. I need to compare it with the existing simulation_test and then will send out an issue summarizing how it would work for review, and if there is interest in a PR, I'll submit one.
There was a problem hiding this comment.
@jpbetz thanks! make sense to me.
|
lgtm. defer to @heyitsanthony |
This is aimed at increasing coverage of
freelist.releaseRangefrom thetxlevel, with a focus on exercising the rollback logic for cases where long running read transactions are holding pages in the pending state.