tree 74e71e22e002607b29658d87b50bf8b9e3d744ca
parent a8d384efca6fdf1d7b542e9e30900fe3a12346f5
author Ben Clayton <bclayton@google.com> 1706886055 +0000
committer Dawn LUCI CQ <dawn-scoped@luci-project-accounts.iam.gserviceaccount.com> 1706886055 +0000

[tools][cts] Fix tree root calculations

addNewExpectations() attempts to find all the 'root nodes' across
all the variants. These are the points for each variant where we'd put
an expectation. We collect all these root nodes across all the variants,
before building the new expectations.

The logic to do this used the query.Tree.Split() method, documented as:
"Split adds a new data to the tree, clearing any ancestor node's data."

This is not the right method for building the root tree. Consider:

Variant A has failing cases 'a:b,c:d:x,*' and 'a:b,c:d:y,*'
Variant B has failing cases 'a:b,c:d:x,*' ('a:b,c:d:y,*' pass)

Variant A adds the root node:      'a:b,c:d:*'
Variant B then adds the root node: 'a:b,c:d:x,*'

'a:b,c:d:x,*' is a descendant of 'a:b,c:d:*', so the root node for
variant A gets removed, and we don't end up emitting the expectations for Variant A's 'a:b,c:d:y,*'.

To fix this, use query.Tree.Add(). This results in the root node tree
potentially having multiple nodes along a query, but overlaps are
already handled with expectationsForRoot(), so this is fine.

Fixed: tint:2384
Change-Id: Ibeb244940885faa4b093b9039fa466681a5aa560
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/172840
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Auto-Submit: Ben Clayton <bclayton@google.com>