[Merged by Bors] - Only run benchmarks on PRs when a label is set

[raskad]

This changes our ci benchmarks to only run when the label run-benchmark is set on the PR.

The motivation is to reduce the time waiting on benchmarks to run while working on PRs. Also this saves some ci minutes which is always good.
When we spot changes that we suspect impact performance, we can add the run-benchmark label to the PR and the benchmarks will run.

[jedel1043]

Nice! This will definitely speedup our CI process by a lot :)

[github-actions]

Test262 conformance changes

VM implementation

Test result	main count	PR count	difference
Total	90,499	90,499	0
Passed	56,446	56,446	0
Ignored	23,580	23,580	0
Failed	10,473	10,473	0
Panics	0	0	0
Conformance	62.37%	62.37%	0.00%

[codecov]

Codecov Report

Merging #2114 (5e1a1ae) into main (5002b9b) will decrease coverage by 0.03%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##             main    #2114      +/-   ##
==========================================
- Coverage   43.66%   43.62%   -0.04%     
==========================================
  Files         217      217              
  Lines       19644    19673      +29     
==========================================
+ Hits         8578     8583       +5     
- Misses      11066    11090      +24     

Impacted Files	Coverage Δ
boa_engine/src/syntax/ast/node/spread/mod.rs	57.14% <0.00%> (-14.29%)	⬇️
boa_engine/src/builtins/object/mod.rs	59.91% <0.00%> (-6.75%)	⬇️
...x/ast/node/declaration/async_generator_decl/mod.rs	6.66% <0.00%> (-6.67%)	⬇️
...ntax/parser/expression/left_hand_side/arguments.rs	35.00% <0.00%> (-2.50%)	⬇️
boa_engine/src/vm/code_block.rs	45.81% <0.00%> (ø)
boa_engine/src/syntax/ast/node/mod.rs	48.85% <0.00%> (ø)
boa_engine/src/syntax/ast/node/new/mod.rs	37.50% <0.00%> (ø)
boa_engine/src/syntax/ast/node/call/mod.rs	69.23% <0.00%> (ø)
boa_engine/src/syntax/ast/node/array/mod.rs	33.33% <0.00%> (ø)
boa_engine/src/syntax/ast/node/block/mod.rs	50.00% <0.00%> (ø)
... and 38 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5002b9b...5e1a1ae. Read the comment docs.

[jedel1043]

I saw the test conformance comment and it occured to me that it would be good to conditionally run parts of the CI only when there has been changes on boa_engine, boa_gc, boa_interner, etc

[raskad]

I saw the test conformance comment and it occured to me that it would be good to conditionally run parts of the CI only when there has been changes on boa_engine, boa_gc, boa_interner, etc

True. I think there is some discussion needed on this, as some changes outside of the Rust crates probably also have to be tested. For example if we update the 262 submodule, we should probably run ci normally.

[Razican]

The thing is that a change to, for example, the Unicode dependency, could make some random stuff to fail, and performance to change in any place (almost). So it's difficult to know when it makes sense or not.

[raskad]

The thing is that a change to, for example, the Unicode dependency, could make some random stuff to fail, and performance to change in any place (almost). So it's difficult to know when it makes sense or not.

That is true, but I would argue that those cases should not be the norm and we still have the full benchmarks on main. If any change unexpectedly changes performance, we can still track it down to the commit. IMO the benefits of getting faster feedback on PRs outweigh the negatives.

[jasonwilliams]

I’m assuming this works at the point when you add the label?

[raskad]

I’m assuming this works at the point when you add the label?

Yes, when you add the label it gets triggered for the first time. Any push after that also triggers them like in the past.

[Razican]

Looks good to me :)

[Razican]

bors r+

[bors]

Pull request successfully merged into main.

Build succeeded:

• Clippy
• Documentation
• Examples
• Rustfmt
• Tests on Linux
• Tests on MacOS
• Tests on Windows