Parameterise the generated file's licence and copyright

[thedavidmeister]

Closes #75

LibCodeGen.filePrefix() returned a hardcoded LicenseRef-DCL-1.0 +
Copyright (c) 2020 Rain Open Source Software Ltd, and
LibFs.buildFileForContract stamped it onto every file it writes, in every
consumer repo, in any org. S01-Issuer/st0x.deploy is live on this library:
its ten src/generated/0_1_1/*.pointers.sol files — append-only deploy-pin
snapshots, the release record consumers pin addresses and codehashes against —
each open with Rain's licence and Rain's copyright holder. reuse lint there
passes, because it checks the tag is present, not that it is right.

What changed

• filePrefix() → filePrefix(string spdxLicenseIdentifier, string copyrightText).
  Both are interpolated verbatim into their own SPDX tag. The pragma and the
  autogenerated notice stay the library's: those are facts about the code it
  emits, not claims about who owns the repo the file lands in.
• New LibCodeGen.isSingleLine(string) and two errors,
  InvalidSpdxLicenseIdentifier(string) and InvalidCopyrightText(string).
  Each value must be non-empty and free of 0x0A/0x0D; the licence is
  checked first.
• LibFs.buildFileForContract takes both and passes them through.

What the merge with main changed in this PR

origin/main at dbcd5e1 is merged in. main moved three times while this was
being finished — #100, then #136, then #127 and #123 together — and three of
those touch the function these parameters run through, so each round is a real
resolution rather than a formality. Seven resolutions that are not mechanical,
and one branch head absorbed rather than taken.

#100 landed, and it rebuilt the function these parameters run through.
buildFileForContract now computes the whole file content into a local before
it touches disk:

string memory content = string.concat(
    LibCodeGen.filePrefix(spdxLicenseIdentifier, copyrightText),
    LibCodeGen.bytecodeHashConstantString(vm, instance),
    body
);
vm.createDir(dir, true);
while (isPresent(vm, path)) { vm.removeFile(path); }
vm.writeFile(path, content);

Resolved onto that ordering rather than restoring this branch's
compute-at-vm.writeFile. That inverts the "Decisions" entry this PR used to
carry, which is rewritten below. #100's own
testBuildFileForContractFailedBuildKeepsExistingFile is kept, passes the two
values like every other call site, and passes — it is the proof the resolution
is the right way round. The docstring paragraph #100 added is widened by one
clause, from "building it reverts for an instance that holds no code" to
"...and for a licence or copyright filePrefix refuses", because this branch is
what adds the second revert path to the step that paragraph is about.

#127 landed too, and turned the unlink into a loop. if (isPresent(...))
became while (isPresent(...)), in the same function, after the content local.
Taken as main has it; nothing about the two threaded values interacts with how
many times the path is unlinked, and #127's own docstring paragraphs about the
symlink cases sit alongside the licence paragraph this PR adds rather than in it.

#127 also added testBuildFileForContractReplacesLiveSymlink, and git merged
it silently with the wrong arity. It calls buildFileForContract twice in
test/src/lib/LibFs.isPresent.t.sol; both sides of the merge left those lines
alone, so there was no conflict marker on them and the tree compiled to
Error (9582): Member "buildFileForContract" not found or not visible after argument-dependent lookup. Threaded, with the same two-line comment the file's
other generating test already carries, saying this repo's own values stand in
where nothing reads the header. This is the whole reason the suite is run on the
merge commit rather than on either parent.

#123 renamed requireContractName to requireIdentifier and
InvalidContractName to InvalidIdentifier. Conflicted with this branch's
import block, which had grown the two new error names. Both sides taken:
{InvalidIdentifier, InvalidSpdxLicenseIdentifier, InvalidCopyrightText, CodelessInstance}. assertNameRejected takes main's new selector and this
branch's argument list.

The remote branch head at 5c68149 is absorbed with git merge -s ours, not
force-pushed over. It was an automated merge of the same origin/main@c8440fb
under a different resolution; it adds no commit this branch lacks. Two things
make this tree the one to keep: 5c68149 restores testFilePrefixNamesNoScript,
which is exactly the test the reasoning below says is subsumed, and it predates
the reorder of the two tests this PR owns to read → clean up → assert. The
ours-merge keeps it in the history and its reason in the commit message rather
than discarding it silently.

#126's forge-std prefix rewrite was a textual conflict. Both sides
edited the import block of test/src/lib/LibCodeGen.filePrefix.t.sol: main
moved Test from forge-std-1.16.1 to forge-std-1.16.2, this branch added the
two error imports and opened the REUSE-ignore region. Both sides are taken.
forge-std-1.16.1 appears nowhere in the tree outside dependencies/, out/
and cache/.

#112's restructure is where the parameters go now. buildFileForContract
is a four-argument wrapper delegating to a five-argument overload taking dir,
and the overload is where the content build, the unlink and the vm.writeFile
live. The two values
are threaded through both — the overload takes seven arguments and the wrapper
six — rather than through the flat function this branch was written against.
testBuildFileForContractCreatesTheDirectory, added on main by #112 and
driving the overload directly, passes them too.

#138 deleted script/Build.sol and src/generated/CodeGennable.sol. The
hunk that made the script pass this repo's own values was a modify/delete; the
deletion is taken and the library now has no in-repo caller.
testBuildFileForContractCommittedArtifactIsCurrent went with them — it read
the deleted artifact — so the two file-scoped constants it existed to pin,
SPDX_LICENSE_IDENTIFIER and COPYRIGHT_TEXT, are now documented as what they
actually are: this repo's own values standing in wherever a test does not care
which values it passes.

#56 moved every .t.sol into the test/src/ mirror. Both test files this
PR changes moved with it; the changes are on the new paths and no old path is
recreated.

testFilePrefixNamesNoScript is not restored. main dropped it in
95a1e89 as implied by testFilePrefixExact in the same file, and main's
docstring for testFilePrefixExact, which carries that reason, is what this
branch now has — with one clause added to say the pinned call passes this repo's
own values as the caller's, which is the only thing parameterisation changes
about it.

Every file this branch touches was diffed against main's version after the
merge. test/src/lib/LibFs.isPresent.t.sol differs from main by exactly the
four threaded call sites — two in the dangling-symlink test, two in #127's new
live-symlink test — and the same two-line comment above each pair saying why
this repo's own values stand in there. Nothing else in that file moves.

Decisions

Required parameters, not an overload keeping a Rain default. A default is
exactly the silent-wrong-header this finding is about: every consumer that never
looks at the new argument keeps emitting Rain's copyright and keeps passing
reuse lint. Required parameters make the compile error the migration notice.
This is also what #75's own proposed fix is — it shows a two-parameter
filePrefix with no default, and says that a single global header leaves the
finding unresolved until something asserts the consumer's repo licence agrees
with it. Ledger — build cost: this diff, plus two arguments at each of the 26
consumer call sites enumerated below, at the version each consumer chooses to
bump. Carrying cost: two extra arguments on a function that is already called
from a build script that has the values in front of it. Cost of removing it
later: reverting to a default would silently re-stamp Rain's holder on whatever
had not been migrated, so it is a one-way door in the correct direction.

Validation is in scope, and the issue's snippet has none. Measured, not
reasoned: a generated file whose SPDX tag has no value is refused by solc with
Error (1114): Invalid SPDX license identifier, so an empty value does not
produce a licensed file, it produces a repo that stops compiling — after the file
has been written and committed. A value carrying \n ends the tag's comment line
and writes a second, contradicting SPDX tag one line down, which a presence check
accepts. Both are the same "present but wrong" failure the finding names, one
level up, so the two guards ship with the parameters rather than after them.
0x0D counts with 0x0A: solc ends a // comment at either.

The revert lands before anything on disk is touched — #100 closed the window
this PR used to argue for leaving open. The earlier version of this body argued
that a refused licence reverts with the path already unlinked, that the window
was not new because LibCodeGen.bytecodeHashConstantString's CodelessInstance
already reverted from the same argument list, and that closing it was #61's job,
in flight as #100. #100 has since merged, so there is no window left to argue
about: filePrefix is called building the content local, which happens before
vm.createDir, before the unlink and before the write. A refused licence or
copyright now leaves the directory exactly as it found it, on the same path
CodelessInstance takes. Nothing was hoisted to get this — the guards are where
they always were, inside filePrefix; #100 moved the call site out in front of
the disk mutation and both revert paths came with it.

testBuildFileForContractRejectsInvalidLicenceAndCopyright asserted
assertFalse(written, "a refused header still wrote a file") under the old
ordering and asserts it under the new one. It held before because the write was
never reached; it holds now because nothing at all is reached. The assertion did
not need changing, which is the useful signal: it was written against the
property, not against the ordering.

One //REUSE-IgnoreStart / //REUSE-IgnoreEnd region in
test/src/lib/LibCodeGen.filePrefix.t.sol, opened after the imports and closed
at EOF. The tag prefixes now appear in source without a value attached, and the
whole subject of that file is emitted licence text, so per-assertion regions
would be false precision. It opens below the file's own header, so that header
still registers — reuse lint reports Files with license information: 51 / 51.
Measured on the merge commit by deleting the two markers and running
nix develop -c reuse lint: exit 1, Invalid SPDX License Expressions: 3, the
three being ", spdxLicenseIdentifier));, GPL-3.0-only" and
LicenseRef-DCL-1.0\n" — fragments of test source that happen to follow an
SPDX-License-Identifier: prefix. With the markers, exit 0 and
Invalid SPDX License Expressions: 0. Files with license information: 51 / 51
either way, which is the point of opening the region below the header rather
than above it.

Failing before, passing after

The change has two halves and each has its own red, both measured on the merge
commit.

The signature. src/lib/LibCodeGen.sol restored to main's version with the new
tests and everything else at HEAD, nix develop -c forge test:

Compiling 22 files with Solc 0.8.25
Solc 0.8.25 finished in 173.11ms
Error: Compiler run failed:
Error (2904): Declaration "InvalidSpdxLicenseIdentifier" not found in "src/lib/LibCodeGen.sol" (referenced as "src/lib/LibCodeGen.sol").
 --> test/src/lib/LibCodeGen.filePrefix.t.sol:6:1:
  |
6 | import {LibCodeGen, InvalidSpdxLicenseIdentifier, InvalidCopyrightText} from "src/lib/LibCodeGen.sol";
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Error (2904): Declaration "InvalidCopyrightText" not found in "src/lib/LibCodeGen.sol" (referenced as "src/lib/LibCodeGen.sol").
 --> test/src/lib/LibCodeGen.filePrefix.t.sol:6:1:
  |
6 | import {LibCodeGen, InvalidSpdxLicenseIdentifier, InvalidCopyrightText} from "src/lib/LibCodeGen.sol";
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

The interpolation. Parameters present and ignored — the body still hardcoding
Rain's licence, which is M01 below:

Ran 23 test suites in 3.33s (34.61s CPU time): 182 tests passed, 3 failed, 0 skipped (185 total tests)

Failing tests:
Encountered 2 failing tests in test/src/lib/LibCodeGen.filePrefix.t.sol:LibCodeGenFilePrefixTest
[FAIL: assertion failed: // SPDX-License-Identifier: LicenseRef-DCL-1.0
// SPDX-FileCopyrightText: Copyright (c) 2026 S01 Issuer GmbH
[…] != // SPDX-License-Identifier: MIT
// SPDX-FileCopyrightText: Copyright (c) 2026 S01 Issuer GmbH
[…] testFilePrefixCarriesTheCallersLicenceAndCopyright() (gas: 11786)
[FAIL: assertion failed: // SPDX-License-Identifier: LicenseRef-DCL-1.0 != // SPDX-License-Identifier: X-r|$wQ_o lPgW?nx60_Y; counterexample: […]] testFilePrefixLinesAreTheCallersValues(bytes,bytes) (runs: 0, μ: 0, ~: 0)

Encountered 1 failing test in test/src/lib/LibFs.buildFileForContract.t.sol:LibFsBuildFileForContractTest
[FAIL: assertion failed: // SPDX-License-Identifier: LicenseRef-DCL-1.0
// SPDX-FileCopyrightText: Copyright (c) 2026 S01 Issuer GmbH
[…] != // SPDX-License-Identifier: MIT
// SPDX-FileCopyrightText: Copyright (c) 2026 S01 Issuer GmbH
[…] testBuildFileForContractCarriesTheCallersLicenceAndCopyright() (gas: 103002)

Encountered a total of 3 failing tests, 182 tests succeeded

[…] elides the rest of each assertion's two whole file contents and the fuzz
counterexample's calldata. Three tests fail and 182 pass: the caller's licence
does not reach the header, at both levels, and neither the exact-prefix pin nor
any other test in the suite notices — which is why the two "carries the caller's
values" tests exist. The copyright half is symmetric and is M02.

After the fix, nix develop -c forge test reports
185 tests passed, 0 failed, 0 skipped (185 total tests).

Mutation matrix

Run with mutation-probe, over the pushed tree at 63f727a — the head this PR
carries, not an earlier merge. It refuses to score anything until the baseline is
green (185 passed), requires each mutant's target text to occur exactly once in
its file, and scores a run with no tally from forge as NO-RUN rather than as a
survivor. cache/fuzz/failures is deleted before every run, so no verdict rests
on a counterexample replayed from a previous mutant.

Re-run in full after each main merge, because #100 and #127 both rewrote the
body of the function M10 and M13 mutate. Both target strings survived those
rewrites unchanged: LibCodeGen.filePrefix(spdxLicenseIdentifier, copyrightText),
moved from inside the vm.writeFile call to inside the content local and kept
its trailing comma, and the wrapper's delegation line was never touched. So all
13 are real runs, not silently skipped ones. That is the thing to check rather
than assume — a mutant whose target text no longer occurs does not run at all,
and reporting it as KILLED would be a matrix that lies. What rules it out here
is the probe's own "target must occur exactly once" precondition and the
no-run: 0 in the tally below, not reading the diff and hoping.

Mutant	File	Verdict	Killed by
M01 filePrefix interpolates the literal LicenseRef-DCL-1.0 instead of spdxLicenseIdentifier	LibCodeGen.sol	KILLED	testBuildFileForContractCarriesTheCallersLicenceAndCopyright, testFilePrefixCarriesTheCallersLicenceAndCopyright, testFilePrefixLinesAreTheCallersValues
M02 filePrefix interpolates the literal Rain copyright instead of copyrightText	LibCodeGen.sol	KILLED	testBuildFileForContractCarriesTheCallersLicenceAndCopyright, testFilePrefixCarriesTheCallersLicenceAndCopyright, testFilePrefixLinesAreTheCallersValues
M03 isSingleLine stops rejecting the empty string	LibCodeGen.sol	KILLED	testBuildFileForContractRejectsInvalidLicenceAndCopyright, testFilePrefixChecksLicenceFirst, testFilePrefixRejectsEmptyCopyright, testFilePrefixRejectsEmptyLicence
M04 isSingleLine stops rejecting 0x0A	LibCodeGen.sol	KILLED	testBuildFileForContractRejectsInvalidLicenceAndCopyright, testFilePrefixRejectsLineBreakAtAnyPosition, testFilePrefixRejectsLineBreakInCopyright, testFilePrefixRejectsLineBreakInLicence
M05 isSingleLine stops rejecting 0x0D	LibCodeGen.sol	KILLED	testFilePrefixRejectsLineBreakAtAnyPosition, testFilePrefixRejectsLineBreakInCopyright, testFilePrefixRejectsLineBreakInLicence
M06 isSingleLine never inspects the last byte	LibCodeGen.sol	KILLED	testBuildFileForContractRejectsInvalidLicenceAndCopyright, testFilePrefixRejectsLineBreakAtAnyPosition, testFilePrefixRejectsLineBreakInCopyright, testFilePrefixRejectsLineBreakInLicence
M07 filePrefix drops the InvalidSpdxLicenseIdentifier guard	LibCodeGen.sol	KILLED	testBuildFileForContractRejectsInvalidLicenceAndCopyright, testFilePrefixChecksLicenceFirst, testFilePrefixRejectsEmptyLicence, testFilePrefixRejectsLineBreakAtAnyPosition, testFilePrefixRejectsLineBreakInLicence
M08 filePrefix drops the InvalidCopyrightText guard	LibCodeGen.sol	KILLED	testBuildFileForContractRejectsInvalidLicenceAndCopyright, testFilePrefixRejectsEmptyCopyright, testFilePrefixRejectsLineBreakAtAnyPosition, testFilePrefixRejectsLineBreakInCopyright
M09 filePrefix checks the copyright before the licence	LibCodeGen.sol	KILLED	testFilePrefixChecksLicenceFirst
M10 buildFileForContract passes the two values to filePrefix in the wrong order	LibFs.sol	KILLED	testBuildFileForContractBodyVerbatim, testBuildFileForContractCarriesTheCallersLicenceAndCopyright, testBuildFileForContractCreatesTheDirectory, testBuildFileForContractEmptyBody, testBuildFileForContractExactContent
M11 filePrefix swaps the two SPDX tag names	LibCodeGen.sol	KILLED	testBuildFileForContractBodyVerbatim, testBuildFileForContractCarriesTheCallersLicenceAndCopyright, testBuildFileForContractCreatesTheDirectory, testBuildFileForContractEmptyBody, testBuildFileForContractExactContent
M12 isSingleLine never inspects the first byte	LibCodeGen.sol	KILLED	testFilePrefixRejectsLineBreakAtAnyPosition
M13 the four-argument wrapper hands the two values to the overload in the wrong order	LibFs.sol	KILLED	testBuildFileForContractBodyVerbatim, testBuildFileForContractCarriesTheCallersLicenceAndCopyright, testBuildFileForContractEmptyBody, testBuildFileForContractExactContent, testBuildFileForContractFreshPath

13/13 killed; survived: 0; no-run: 0; harness errors: 0.

Every name above is the probe's own killed_by list. The probe stops after the
first five names it extracts from a red run, so a five-name row is a witness
list, not the whole set of tests that kill that mutant. The fail-pattern is
anchored on [FAIL and matches lazily across newlines, so a fuzz test — whose
name forge prints after a bracket containing both args=[…] and the
counterexample — is named like any other, and a [PASS] line can never be
mistaken for a killer.

The test that could poison the next compile, and does not any more

The first matrix run of this work stopped producing results at M04 and produced
neither a summary line nor a compilation error — the signature of a harness that
is lying. Reproduced and diagnosed rather than worked around:
testBuildFileForContractRejectsInvalidLicenceAndCopyright used
vm.expectRevert, which ends the test at the call that is supposed to revert and
does not. Under M03 that call succeeds, vm.writeFile puts
src/generated/LibFsBuildInvalidHeader.sol on disk with an empty SPDX tag, and
the revert that expectRevert then synthesises does not undo a filesystem write.
The file survived git checkout -- src test, and every later run died at
Error (1114): Invalid SPDX license identifier before a single test ran.

Fixed in the test: the call is made through try/catch so control comes back
either way, every file is removed, and only then does anything assert. The revert
data is compared whole — selector and arguments — so nothing is weakened relative
to vm.expectRevert.

The same shape bit testBuildFileForContractCarriesTheCallersLicenceAndCopyright
in this run: under M01 it asserts before it cleans up, so
src/generated/LibFsBuildForeignLicence.sol was left behind by the pass. That
file is valid Solidity with a well-formed MIT tag, so it poisons nothing, but
it is untracked residue from a suite that is supposed to leave the tree as it
found it. That test now reads the file, removes it, and asserts afterwards.

Breaking change

14 repos pin rain-sol-codegen: rain.deploy 0.1.6, S01-Issuer/st0x.deploy
0.1.3 (issue #75 says 0.1.0; that is stale), and the other twelve at 0.1.0. Read
from each repo's foundry.toml on its default branch, across every repo in
rainlanguage and S01-Issuer.

Nothing breaks on merge. Soldeer pins are exact and every import carries the
version in its path prefix (rain-sol-codegen-0.1.0/src/lib/LibCodeGen.sol), so
an unbumped repo keeps resolving the zip it pinned. Each repo breaks only when it
chooses to bump, and migrates by forge soldeer update, rewriting the versioned
import prefixes, and passing its own SPDX identifier and copyright — which its
REUSE.toml and LICENSES/ already state.

Call sites that will need the two arguments, counted in each repo's default
branch:

Repo	File	Calls
rainlanguage/rainlang	script/Build.sol	6
rainlanguage/raindex	script/Build.sol	6
rainlanguage/dvin.deploy	script/BuildPointers.sol	3
rainlanguage/rain.verify	script/BuildPointers.sol	2
rainlanguage/rain.deploy	src/lib/LibRainDeploySnapshot.sol	1 + 2 filePrefix()
rainlanguage/rain.flare	script/Build.sol	1
rainlanguage/rain.merkle	script/Build.sol	1
rainlanguage/rain.dia	script/Build.sol	1
rainlanguage/rain.erc4626.words	script/Build.sol	1
rainlanguage/rain.pyth	script/Build.sol	1
rainlanguage/rain.math.float	script/BuildPointers.sol	1
rainlanguage/rain.factory.deploy	script/BuildPointers.sol	1
S01-Issuer/st0x.deploy	script/BuildPointers.sol	1
rainlanguage/rainlang.interface	—	0

The cascade runs through rain.deploy. It is itself a published package that
calls this API from src/lib/LibRainDeploySnapshot.sol — buildFileForContract
once and filePrefix() twice — and 15 repos pin rain-deploy. It has to
decide whether to take the two values as its own parameters or hardcode Rain's;
until it threads them through, st0x.deploy cannot reach a correct header no
matter what it passes — its ten committed pointer files are written through
LibRainDeploySnapshot, not through buildFileForContract. This PR is the first
half of that. No consumer repo is touched here.

Version step

This is a breaking API change and wants a minor step, and the version is not
edited in this PR. It is expressible but not through a workflow input:
rainix-autopublish.yaml's level reaches only the cargo path, while the soldeer
path publishes [package].version verbatim and derives the next as bump_patch
(rainix-static/src/soldeer_gate.rs), whose only invariant is
ver_gt(local, published). So 0.2.0 is a one-line hand edit to foundry.toml
that the gate accepts: it publishes 0.2.0, tags sol-v0.2.0, and bumps to
0.2.1. Left out deliberately — it is a release decision that belongs with the
merge, main is at 0.1.31 and moves on every merge, and every sibling PR in
flight would conflict on that line.

Found, not touched

S01-Issuer/st0x.deploy does not agree with itself about its own copyright: its
REUSE.toml and 68 files under src/ say Rain Open Source Software Ltd, while 6
— including src/concrete/ST0xOrchestrator.sol — say
Copyright (c) 2026 S01 Issuer GmbH. Parameterising here does not settle that;
someone has to decide what that repo declares before it can pass the right value.

isSingleLine accepts more than testFilePrefixAcceptsEveryNonLineBreakValue
reaches, and the name does not say so. The guard refuses only 0x0A and
0x0D, so it accepts NUL, tab, the rest of the C0 controls, 0x7F and every
byte 0x80–0xFF. The test reaches only 0x20–0x7E, because textFromSeed
has to produce something the string cheatcodes can read back and arbitrary bytes
are not valid UTF-8. Whether the guard should refuse the other control
characters — a NUL or a tab in an SPDX tag is not a line break but is not a
usable licence identifier either — is a design decision, not a test gap, and is
left for a human. Widening the test's domain without settling that would only pin
the current behaviour harder.

Every generating test in LibFs.buildFileForContract.t.sol asserts before it
cleans up, so any of them leaves its file in src/generated/ when it fails.
That is pre-existing and repo-wide, not introduced here; only the two tests this
PR adds are reordered, because those are the ones it owns. Visible in the matrix
run above: after 13 mutants the tree held nine untracked files under
src/generated/, all of them from tests that predate this PR
(LibFsBuildIdempotent.sol, LibFsBuildEmptyBody.sol, LibFsBuildOverwrite.sol,
LibFsBuildExact.sol, LibFsBuildInstance.sol, LibFsBuildBodyVerbatim.sol,
LibFsBuildFresh.sol, LibFsBuildSiblingA.sol, LibFsBuildSiblingB.sol), and
neither LibFsBuildForeignLicence.sol nor LibFsBuildInvalidHeader.sol — the
two this PR owns and reordered.

QA

• Discriminating tests: testFilePrefixCarriesTheCallersLicenceAndCopyright and
  testBuildFileForContractCarriesTheCallersLicenceAndCopyright pin a foreign
  caller's values through to the header and assert this repo's own licence and
  copyright holder appear nowhere in the result;
  testFilePrefixLinesAreTheCallersValues fuzzes both values and reads the
  emitted prefix by splitting it on newlines, so tag text, tag order and line
  count are all pinned; testFilePrefixRejects{EmptyLicence,EmptyCopyright, LineBreakInLicence,LineBreakInCopyright,LineBreakAtAnyPosition} and
  testFilePrefixChecksLicenceFirst pin the guards and their order;
  testFilePrefixAcceptsEveryNonLineBreakValue pins that a printable-ASCII
  value is not refused — its name overclaims relative to that domain, see
  "Found, not touched"; testBuildFileForContractRejectsInvalidLicenceAndCopyright
  pins that the write inherits all of it and leaves no file behind.
• Mutations applied: 13, listed with verdicts and killers in the matrix above —
  13/13 killed; survived: 0; no-run: 0; harness errors: 0, over a baseline the
  probe proved green at 185 passed before probing anything. M13 is new to this
  merge: it covers the wrapper-to-overload delegation that Cover the generated directory being created, by making the directory injectable #112's restructure
  created, which did not exist when this branch was written. The two documented
  ways a mutation run lies in this repo are both closed: cache/fuzz/failures is
  deleted before every run so no verdict rests on a replayed counterexample, and
  a run that produces no tally is scored NO-RUN rather than SURVIVED.
• Oracle: the consumer repo's own REUSE.toml and LICENSES/, which is what the
  header has to agree with, plus solc's own SPDX rule for the empty case —
  Error (1114): Invalid SPDX license identifier, observed on a real generated
  file, not reasoned from the spec. The tests read the library's output rather
  than restate its implementation: the prefix is read with vm.split rather than
  rebuilt with the string.concat the library uses.
• Category check: filePrefix() stamps this org's licence and copyright into other repos' generated files #75 asks that a generated file stop claiming this org's licence
  and copyright in repos that are not this org's. Covered at both levels
  (filePrefix and buildFileForContract), for both values, with validation so
  that "parameterised" cannot mean "parameterised into a tag that says nothing".
  The consumer census above is the rest of the answer — nothing outside this repo
  is touched, and foundry.toml is deliberately not bumped.
• Suite, on the merge commit: nix develop -c forge test reports
  185 tests passed, 0 failed, 0 skipped (185 total tests) in 23 suites, against
  174 tests passed, 0 failed, 0 skipped (174 total tests) in 23 suites on
  main at dbcd5e1, measured the same way — 11 added, none lost.
  nix develop -c forge fmt --check exits 0 with no output.
  nix develop -c reuse lint exits 0 with Invalid SPDX License Expressions: 0
  and 51 / 51 files carrying both copyright and licence information.
  nix develop -c forge coverage reports src/lib at 100% lines (107/107),
  statements (125/125), branches (14/14) and functions (21/21).
• Not re-run, because the tree it ran against is gone: the old QA cited
  forge script script/Build.sol regenerating src/generated/CodeGennable.sol
  with an empty git status --porcelain afterwards. Strip the script/Build.sol worked example, leave the library lean #138 deleted both. There is
  no in-repo generation to check for idempotence any more; what replaces it is
  testBuildFileForContractIdempotent, which asserts a second call over the same
  inputs produces the same bytes.

[coderabbitai]

Warning

Review limit reached

@thedavidmeister, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 28 minutes

Limit details: You’ve used all 1 included review currently available under your plan.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 9c84ef5b-18d5-4234-81dd-1452da11e817

📥 Commits

Reviewing files that changed from the base of the PR and between dbcd5e1 and 63f727a.

📒 Files selected for processing (6)

• src/lib/LibCodeGen.sol
• src/lib/LibFs.sol
• test/concrete/LibFsExternal.sol
• test/src/lib/LibCodeGen.filePrefix.t.sol
• test/src/lib/LibFs.buildFileForContract.t.sol
• test/src/lib/LibFs.isPresent.t.sol

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[thedavidmeister]

@coderabbitai review

[coderabbitai]

⚠️ Action not completed

Review rate limited.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.