When comparing SVG tests against their blessed version, ignore the first line

[estebank]

anstyle_svg has some weird non-determinism in the width parameter, which makes tests blessed in one environment to fail in another. This is the only non-determinism detected so far, so we modify the diff check to ignore the first line of the SVG. In order for a test to fail/be updated by --bless, a different part of the file needs to also have changed. If other sources of non-determinism are found, we should back out of the "--color=always means .svg" change.

r? @compiler-errors

[rustbot]

Some changes occurred in src/tools/compiletest

cc @jieyouxu

[estebank]

Context #122029 (comment)

[lqd]

I can reproduce the failure and this does fix it 👍

[compiler-errors]

@bors r+ rollup p=1

[bors]

📌 Commit bf63f7e has been approved by compiler-errors

It is now in the queue for this repository.

[klensy]

I can reproduce the failure and this does fix it 👍

what command?
python x.py --stage=1 --jobs=4 t tests/ui/codemap_tests --force-rerun can't repro on x86_64-pc-windows-msvc

ahh ignore-windows
But on windows still can't catch 750\751 px difference, only color diff.

[bors]

⌛ Testing commit bf63f7e with merge e3df96c...

[lqd]

what command?

A simple ./x test tests/ui/codemap_tests --force-rerun on my x86_64-unknown-linux-gnu.

[estebank]

ahh ignore-windows
But on windows still can't catch 750\751 px difference, only color diff.

The distinction is made because on windows we produce different colors than on all other platforms (cyan instead of light blue), which would require everyone that works with ui tests to have access to both a linux/mac and windows, in perpetuity.

[bors]

☀️ Test successful - checks-actions
Approved by: compiler-errors
Pushing e3df96c to master...

[rust-timer]

Finished benchmarking commit (e3df96c): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.7%	[-2.7%, -2.7%]	1
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	6.7%	[6.6%, 6.9%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.7%	[2.7%, 2.7%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 669.176s -> 669.218s (0.01%)
Artifact size: 312.73 MiB -> 312.87 MiB (0.05%)

[klensy]

what command?

A simple ./x test tests/ui/codemap_tests --force-rerun on my x86_64-unknown-linux-gnu.

Still all ok, strange.

[estebank]

At work we noticed that this fix didn't make it into 1.78, which makes the .svg tests fail non-deterministically. If we do a dot-release, we should include this PR.

[apiraino]

Stable backport accepted (if a dot release will be planned) as per compiler team on Zulip. A backport PR will be authored by the release team at the end of the current development cycle.

@rustbot label +stable-accepted