693431 - dense array specialized fast-path white-box tests

Status: RESOLVED WONTFIX

(Tamarin Graveyard :: Library, defect, P3)

Description

[Felix S. Klock II [:pnkfelix, :fklock]]

A number of bugs (tracked by Bug 690498) have gone undetected because the dense array rewrite introduced several new control paths but those paths were not properly exercised by our test suite.

We need a set of glass box tests for this code.

I have a patch in progress.  I believe it has uncovered two bugs that were otherwise undetected until now.  I hope to post the test shortly.

Comment 1

[Dan Smith]

Bravo!

Comment 2

[Felix S. Klock II [:pnkfelix, :fklock]]

Attachment: patch W v1: whitebox tests

(not asking for feedback or review yet.  there's a follow up bug I want to file and try to fix before putting this through the review cycle.)

Comment 3

[Felix S. Klock II [:pnkfelix, :fklock]]

Attachment: patch W v2: whitebox tests

Revised comments/documentation.

Added separate set of tests to cover holes introduced by delete (as opposed to holes introduced by non-contiguous assignments).

Made the boolean flag controlling whether we are including strict array comparison a global constant in the test.

Comment 4

[Felix S. Klock II [:pnkfelix, :fklock]]

Attachment: patch W v3: whitebox tests

* Added more initial arrays

* Added more test operations

* Added timing instrumentation

* Narrowed down a couple of the tests to not create such large arrays (to speed up operations that take time proportional to array length)

Comment 5

[Felix S. Klock II [:pnkfelix, :fklock]]

Attachment: patch W v4: whitebox tests

Added type annotations and uncondtional traversal of indices 0..9 in response to Bug 702262, comment 3.

Comment 6

[Felix S. Klock II [:pnkfelix, :fklock]]

Attachment: patch W v5: whitebox tests (in 10 separate parts)

The test became large enough to justify division into separate pieces.

The main benefit here is that it is easier to see where time is going as the test runs.

It also may be easier to read and extract individual failure cases this way.

Comment 7

[Felix S. Klock II [:pnkfelix, :fklock]]

It is somewhat shameful that I never got this to the point of landing it.  I think either brbaker or dschaffe could take it on.  Alternatively, one of the new QE members could probably take it on as a good initial learning task.

But for now I will just unassign myself from it.

Comment 8

[Sylvestre Ledru [:Sylvestre]]

Tamarin isn't maintained anymore. WONTFIX remaining bugs.