fix!: Dagger flags cancelation inside daggarable functions

[nicolaassolini-qntm]

Closes #1984.

This PR fixes an issue where dagger blocks incorrectly cancelled the daggerable property of the enclosing function: a daggerable function should always enforce the daggerable constraint within its body. Previously, a dagger block inside a daggerable function was treated as non-daggerable because the block's dagger modifier cleared the daggerable flag inherited from the function decorator. However, since the context in which a function is called is not known when checking its body, the unitary constraints declared in the function signature must always be enforced.

The bug was generating a bad cfg.unitary_flags in the modifier blocks, affecting both the metadata and the type checking.

BREAKING CHANGE: guppylang-internals: remaned parameter unitary_flags CFGBuilder.build to function_unitary_flags

[hugrbot]

This PR contains breaking changes to the public Python API.

Breaking changes summary

guppylang-internals/src/guppylang_internals/cfg/builder.py:90: CFGBuilder.build(unitary_flags):
Parameter was removed

[github-actions]

Bencher Report

Branch	na/1984-flags
Testbed	Linux

Click to view all benchmark results

Benchmark	hugr_bytes	Benchmark Result bytes x 1e3 (Result Δ%)	Upper Boundary bytes x 1e3 (Limit %)	hugr_nodes	Benchmark Result nodes (Result Δ%)	Upper Boundary nodes (Limit %)
tests/benchmarks/test_big_array.py::test_big_array_compile	📈 view plot 🚷 view threshold	120.67 x 1e3 (0.00%) Baseline: 120.67 x 1e3	121.87 x 1e3 (99.01%)	📈 view plot 🚷 view threshold	5,659.00 (0.00%) Baseline: 5,659.00	5,715.59 (99.01%)
tests/benchmarks/test_ctrl_flow.py::test_many_ctrl_flow_compile	📈 view plot 🚷 view threshold	36.19 x 1e3 (0.00%) Baseline: 36.19 x 1e3	36.55 x 1e3 (99.01%)	📈 view plot 🚷 view threshold	1,582.00 (0.00%) Baseline: 1,582.00	1,597.82 (99.01%)
tests/benchmarks/test_queue_push_pop.py::test_queue_push_benchmark_compile	📈 view plot 🚷 view threshold	9.95 x 1e3 (0.00%) Baseline: 9.95 x 1e3	10.05 x 1e3 (99.01%)	📈 view plot 🚷 view threshold	347.00 (0.00%) Baseline: 347.00	350.47 (99.01%)
tests/benchmarks/test_queue_push_pop.py::test_queue_push_pop_benchmark_compile	📈 view plot 🚷 view threshold	14.11 x 1e3 (0.00%) Baseline: 14.11 x 1e3	14.25 x 1e3 (99.01%)	📈 view plot 🚷 view threshold	470.00 (0.00%) Baseline: 470.00	474.70 (99.01%)

🐰 View full continuous benchmarking report in Bencher

[Copilot]

Pull request overview

This PR fixes unitary-flag inference for modifier (with ...) blocks so that unitary flags declared on a function via the @guppy(...) decorator are preserved inside nested modifier blocks (preventing accidental cancellation via UnitaryFlags.accumulate, e.g. with dagger).

Changes:

• Split function-level decorator unitary flags from block-accumulated unitary flags in CFGBuilder, and ensure decorator flags are preserved when constructing modifier-block CFGs.
• Add a regression test ensuring a @guppy(daggerable=True) function containing a with dagger: block is still treated as daggerable for type checking.
• Add the corresponding expected error output for the new regression test.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

File	Description
tests/error/modifier_errors/decorator_dagger_is_preserved.py	New regression test exercising daggerable decorator preservation through a dagger modifier block.
tests/error/modifier_errors/decorator_dagger_is_preserved.err	Expected error output confirming the dagger-context constraint is enforced.
guppylang-internals/src/guppylang_internals/cfg/builder.py	Adjust CFG construction to track decorator flags separately from accumulated modifier flags and preserve decorator flags in nested blocks.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.02%. Comparing base (f711512) to head (d5ae581).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2003   +/-   ##
=======================================
  Coverage   93.02%   93.02%           
=======================================
  Files         152      152           
  Lines       14609    14615    +6     
=======================================
+ Hits        13590    13596    +6     
  Misses       1019     1019           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[codspeed-hq]

Merging this PR will not alter performance

✅ 11 untouched benchmarks

---

_{Comparing na/1984-flags (d5ae581) with main (f711512)}

[mark-koch]

Wouldn't the easier fix be to just pass self.unitary_flags here instead of accumulated_flags?

[nicolaassolini-qntm]

do you mean self.function_unitary_flags?

[mark-koch]

I mean keep the single unitary_flags field on this class as before and set accumulated_flags = self.unitary_flags | modifiers.flags() here?

[nicolaassolini-qntm]

When there are more nested modifier blocks, I need one variable to store the outer function flag and another one to store the "dynamic" flag of the modifiers.
If I have only one variable, how do I know if self.unitary_flags is related to the function (thus we need to do self.unitary_flags | modifiers.flags()) or to the previous modifier (we need to use the .accumulate method that deletes even daggers)

[mark-koch]

Why do we need to delete even daggers? If I add with dagger, then the body ought to be daggerable imo, even if it's an even dagger

[nicolaassolini-qntm]

with dagger:
  with control:
    with dagger:
       Does the dagger constraint apply here?

[mark-koch]

Imo it should

[mark-koch]

Nevermind, I changed my mind. I think this is ok as is 👍