Emit plain .d.ts for libraries with outputExtensions (TS 7 nightly 20261014)

[NullVoxPopuli]

Important

Still blocked. Two upstream problems remain on 7.1.0-dev.20261015.1. See "Before this can continue" below.

Libraries can now emit counter.d.ts instead of counter.d.gts.ts, so declarations/ matches dist/ file for file. This uses the new outputExtensions option from microsoft/TypeScript#64581, which fixes microsoft/TypeScript#64053.

Changes

• examples/cli-library-v2: sets outputExtensions: { ".gts": ".js", ".gjs": ".js" } and rewriteRelativeImportExtensions: true. The emitted index.d.ts now says from './components/counter.js', and the ./* → ./declarations/*.d.ts export works for every module. I removed the README note saying it didn't.
• README: new "Publishing a library" section. MIGRATING_FROM_TS6.md links the #64053 item to it.
• Test: a new declaration-output-extensions fixture checks that Greeting.d.ts is emitted and Greeting.d.gts.ts is not.
• Bumps every TS 7 pin from 7.1.0-dev.20260821.1 to 7.1.0-dev.20261014.1.

Why the mapper's package.json doesn't set a default

typescript.contentMapper.outputExtensions can go in the mapper's package.json, but I didn't add it:

• Any outputExtensions requires rewriteRelativeImportExtensions. Without it, tsc reports TS18120, even with --noEmit.
• With the flag on, both example apps fail with TS2877. Their extension imports aren't relative (#app/..., cli-app/...), so they can't be rewritten.
• So a default would break every app on upgrade. Apps could opt out with "outputExtensions": {}, but libraries are the ones that emit.

The other manifest fields don't help either: compilerOptions only lists which options get forwarded to the mapper. It can't set them.

Before this can continue

Now on 7.1.0-dev.20261015.1 (second commit). Two problems remain.

@extends over Component.extend(Mixin)

microsoft/TypeScript#64072 applies the tag's type arguments to the call's return type. Ember's Component.extend(Evented) returns a constructor that is not generic.

• The error is now TS2508: No base constructor has the specified number of type arguments, and the signature is still lost.
• Plain-JS repro, no Glint: TS 5.9.3 passes, 20261015.1 fails.
• The upstream-differences snapshot and the docs match the new output. JSDoc @extends signature is lost for classic components with expression heritage (Component.extend(Mixin)) #12 stays open.
• Fix: [BUGFIX] Keep generic classes generic through extend() emberjs/ember.js#21727 keeps the constructor generic. With it, the fixture reports no errors.

Go-to-definition on a .gts import specifier

test/lsp.js "go to definition from a .ts import into a .gts module" fails.

• It returns []. On 20260821.1 it returned counter.gts at 0:0.
• Specifiers for plain .ts files and lookups by name still work.
• Caused by #63936 (Content mappers round 2): last good nightly 20260821.1, first bad 20260822.1.
• No mapper output can fix it. TS needs one span-map segment to cover the whole target file.
• Reported as Content mappers: go-to-definition on an module specifier returns nothing when the target file has more than one span-map segment microsoft/TypeScript#64828.

The outputExtensions work itself is done.

Tracking: #23

🤖 Generated with Claude Code

[github-actions]

🏎️ Benchmark Comparison

	Benchmark	Control (p50)	Experiment (p50)	Δ	Noise
⚪	transform	363.94 ms	362.04 ms	-0.5%	±5.5%
⚪	server 8-in-flight	393.38 ms	373.87 ms	-5.0%	±9.1%
⚪	tsc total	1736.00 ms	1686.00 ms	-2.9%	±3.6%
⚪	tsc mapper wait	1723.00 ms	1672.00 ms	-3.0%	±3.1%

🟢 faster · 🔴 slower · 🟠 slightly slower · ⚪ within 2% or within round-to-round noise

Full mitata output

clk: ~4.34 GHz
cpu: AMD EPYC 9V45 96-Core Processor
runtime: node 24.21.0 (x64-linux)

benchmark                   avg (min … max) p75 / p99    (min … top 1%)
------------------------------------------- -------------------------------
transform (control)          418.61 ms/iter 462.45 ms   █                  
                    (346.27 ms … 541.84 ms) 541.28 ms   ██   █             
                    (844.17 kb …  71.91 mb)  43.10 mb █████▁██▁▁▁▁█▁█▁█▁▁▁█

server 8-in-flight (control) 405.24 ms/iter 412.69 ms        █   █         
                    (382.31 ms … 440.05 ms) 434.64 ms  █     █   █         
                    (  9.68 mb …   9.79 mb)   9.74 mb ████▁▁▁█▁▁▁████▁▁▁▁▁█

━━━ round 1/3: mapper experiment (core 2) ━━━

clk: ~4.33 GHz
cpu: AMD EPYC 9V45 96-Core Processor
runtime: node 24.21.0 (x64-linux)

benchmark                      avg (min … max) p75 / p99    (min … top 1%)
---------------------------------------------- -------------------------------
transform (experiment)          376.57 ms/iter 387.81 ms  █                   
                       (343.59 ms … 490.07 ms) 412.98 ms  ██  █     █ █       
                       (  2.71 mb …  61.64 mb)  42.82 mb ███▁▁██▁▁▁▁█▁█▁▁▁▁█▁█

server 8-in-flight (experiment) 382.51 ms/iter 400.90 ms   █   █             █
                       (339.41 ms … 441.10 ms) 425.52 ms ▅ █ ▅▅█ ▅ ▅ ▅▅▅▅    █
                       (  9.68 mb …   9.78 mb)   9.74 mb █▁█▁███▁█▁█▁████▁▁▁▁█

━━━ round 1/3: tsc control ━━━

tsc: 36 files (6 fixtures × 6), 3 runs after warm-up
  run 1: total 1.83s, mapper wait 1.81s
  run 2: total 1.74s, mapper wait 1.72s
  run 3: total 1.71s, mapper wait 1.69s
  p50: total 1.74s, mapper wait 1.72s

━━━ round 1/3: tsc experiment ━━━

tsc: 36 files (6 fixtures × 6), 3 runs after warm-up
  run 1: total 1.75s, mapper wait 1.74s
  run 2: total 1.72s, mapper wait 1.70s
  run 3: total 1.69s, mapper wait 1.68s
  p50: total 1.72s, mapper wait 1.70s

━━━ round 2/3: mapper control (core 2) ━━━

clk: ~4.37 GHz
cpu: AMD EPYC 9V45 96-Core Processor
runtime: node 24.21.0 (x64-linux)

benchmark                   avg (min … max) p75 / p99    (min … top 1%)
------------------------------------------- -------------------------------
transform (control)          368.02 ms/iter 375.42 ms     ██     █         
                    (346.03 ms … 402.27 ms) 401.25 ms ▅▅▅ ██▅▅ ▅▅█  ▅     ▅
                    (428.09 kb …  72.71 mb)  39.74 mb ███▁████▁███▁▁█▁▁▁▁▁█

server 8-in-flight (control) 379.23 ms/iter 393.90 ms █     █              
                    (350.06 ms … 411.91 ms) 406.15 ms █▅ ▅▅ █   ▅ ▅ ▅▅▅ ▅▅▅
                    (  9.67 mb …   9.96 mb)   9.76 mb ██▁██▁█▁▁▁█▁█▁███▁███

━━━ round 2/3: mapper experiment (core 0) ━━━

clk: ~4.39 GHz
cpu: AMD EPYC 9V45 96-Core Processor
runtime: node 24.21.0 (x64-linux)

benchmark                      avg (min … max) p75 / p99    (min … top 1%)
---------------------------------------------- -------------------------------
transform (experiment)          366.15 ms/iter 371.85 ms       █     █        
                       (334.89 ms … 399.48 ms) 395.99 ms ▅ ▅▅  █▅▅▅▅▅█   ▅  ▅▅
                       (  3.27 mb …  68.09 mb)  39.77 mb █▁██▁▁███████▁▁▁█▁▁██

server 8-in-flight (experiment) 401.99 ms/iter 413.48 ms            ██    █   
                       (362.75 ms … 432.66 ms) 423.93 ms ▅  ▅    ▅▅ ██▅ ▅▅█▅ ▅
                       (  9.68 mb …   9.81 mb)   9.74 mb █▁▁█▁▁▁▁██▁███▁████▁█

━━━ round 2/3: tsc experiment ━━━

tsc: 36 files (6 fixtures × 6), 3 runs after warm-up
  run 1: total 1.68s, mapper wait 1.67s
  run 2: total 1.72s, mapper wait 1.70s
  run 3: total 1.68s, mapper wait 1.66s
  p50: total 1.68s, mapper wait 1.67s

━━━ round 2/3: tsc control ━━━

tsc: 36 files (6 fixtures × 6), 3 runs after warm-up
  run 1: total 1.67s, mapper wait 1.66s
  run 2: total 1.72s, mapper wait 1.70s
  run 3: total 1.72s, mapper wait 1.70s
  p50: total 1.72s, mapper wait 1.70s

━━━ round 3/3: mapper control (core 0) ━━━

clk: ~4.19 GHz
cpu: AMD EPYC 9V45 96-Core Processor
runtime: node 24.21.0 (x64-linux)

benchmark                   avg (min … max) p75 / p99    (min … top 1%)
------------------------------------------- -------------------------------
transform (control)          365.11 ms/iter 371.45 ms   █    █    █        
                    (343.46 ms … 402.48 ms) 383.87 ms ▅▅█    █ ▅▅▅█ ▅ ▅ ▅ ▅
                    (272.66 kb …  59.44 mb)  39.08 mb ███▁▁▁▁█▁████▁█▁█▁█▁█

server 8-in-flight (control) 391.25 ms/iter 405.89 ms          █        █  
                    (356.59 ms … 413.15 ms) 410.84 ms          █       ██  
                    (  9.68 mb …   9.79 mb)   9.75 mb ██▁█▁▁▁▁▁█▁▁▁██▁▁████

━━━ round 3/3: mapper experiment (core 2) ━━━

clk: ~4.18 GHz
cpu: AMD EPYC 9V45 96-Core Processor
runtime: node 24.21.0 (x64-linux)

benchmark                      avg (min … max) p75 / p99    (min … top 1%)
---------------------------------------------- -------------------------------
transform (experiment)          366.61 ms/iter 378.74 ms          █           
                       (333.15 ms … 407.16 ms) 399.24 ms       █  ██          
                       (  5.33 mb …  70.56 mb)  42.38 mb █▁▁█▁▁█████▁▁▁███▁▁▁█

server 8-in-flight (experiment) 369.80 ms/iter 383.32 ms  █                   
                       (348.05 ms … 399.88 ms) 393.23 ms ██               █   
                       (  9.69 mb …   9.86 mb)   9.75 mb ██▁▁▁▁███▁█▁█▁▁███▁▁█

━━━ round 3/3: tsc control ━━━

tsc: 36 files (6 fixtures × 6), 3 runs after warm-up
  run 1: total 1.78s, mapper wait 1.76s
  run 2: total 1.69s, mapper wait 1.67s
  run 3: total 1.78s, mapper wait 1.76s
  p50: total 1.78s, mapper wait 1.76s

━━━ round 3/3: tsc experiment ━━━

tsc: 36 files (6 fixtures × 6), 3 runs after warm-up
  run 1: total 1.70s, mapper wait 1.68s
  run 2: total 1.66s, mapper wait 1.65s
  run 3: total 1.69s, mapper wait 1.67s
  p50: total 1.69s, mapper wait 1.67s