fix(pascal): detect unqualified paren-less calls (DoWork;)

[equipados]

What

A Pascal statement that is nothing but an identifier — DoWork; — is an
unqualified call to a parameterless routine. visitPascalBlock didn't treat it
as one, so only the parenthesised form DoWork() produced a calls edge.

Pascal lets a no-arg routine drop its parens, and in Delphi that is the
convention rather than the exception, so on a real Delphi codebase most
intra-unit calls were invisible to callers / callees / impact. Those tools
didn't return a wrong answer — they returned an empty one, which is harder to
notice.

The qualified form (Obj.Free;) is already handled by
extractPascalParenlessCall (thanks — that landed in 1.5). This is the same
idea one step less qualified.

Measured on a real Delphi codebase

285 .pas/.dpr files from a production Delphi 10.2 ERP (~1.1M LOC):

	calls edges
1.5.0 as-is	2316
with this patch	2711
	+395 (+17%)

How it's scoped

Gated on the identifier being the whole statement:

child.type === 'identifier' &&
node.type === 'statement' &&
node.namedChildCount === 1

Anywhere else — assignment RHS, if condition, any expression — a bare
identifier is indistinguishable from a variable read without type information,
so those are deliberately left alone. That is what the second test locks in:

procedure TFoo.DoThing;
var
  Total, Other: Integer;   // and TFoo has a method also called Total
begin
  Other := Total;                    // variable read, NOT a call
  if Total > 0 then Other := 0;      // variable read, NOT a call
end;

isCalled('TFoo::Total') must stay false.

Tests

Two added next to the existing paren-less test in
__tests__/resolution.test.ts:

• extracts UNQUALIFIED paren-less calls (Reset;, the dominant Delphi idiom) — fails without the change, passes with it.
• does not turn a bare identifier that is NOT a whole statement into a call — the negative case above.

resolution.test.ts + extraction.test.ts: 788 passed. The 5 failures on my
machine (3× PHP include, 1× C/C++ include, 1× gitlink) are identical before and
after the change — pre-existing and unrelated, likely Windows-specific.

tsc --noEmit clean.

Notes

Only src/extraction/tree-sitter.ts changes; Pascal is not part of the Rust
kernel so there is no parity fixture to update.

[danusha2345]

Took this into a local integration build (current main + the open fixes I could merge cleanly) and ran the full suite with the native kernel: 245 files, 4323 passed, 0 failed. No regression on the codegraph repo's own index after a forced re-index. Looks good to me.

[equipados]

Thanks for putting it through a full integration build — the forced re-index on codegraph's own repo is exactly the check I wanted to see, since the risk here was changing edges where the parser already behaved. If it looks good to you, could you submit it as an approving review? The PR still shows REVIEW_REQUIRED, so it can't be merged on the comment alone. Happy to rebase on current main if that helps. Guillermo Álvarez Leal *https://equipados.es* <https://equipados.es> 38611 Granadilla de Abona Tlf:: 655 778 795 Este correo electrónico y, en su caso, cualquier fichero anexo al mismo, contiene información de carácter confidencial exclusivamente dirigida a su destinatario. En caso de su divulgación, copia o distribución a terceros sin la previa autorización escrita de Equipados, usted podría estar vulnerando la legislación vigente. En el caso de haber recibido este correo electrónico por error, se ruega que nos notifique inmediatamente esta circunstancia mediante reenvío a la dirección electrónica del remitente. De conformidad con lo establecido en la LOPD Equipados garantiza la adopción de las medidas necesarias para asegurar el tratamiento confidencial de los datos de carácter personal. Así mismo le informamos de inclusión de sus datos en un fichero con la finalidad de mantener su relación contractual con Equipados. y de la posibilidad de ejercer los derechos de acceso, rectificación, cancelación y oposición en la siguiente dirección: Apartado de correos, nº 190, C.P 38611 – Granadilla de Abona - Santa Cruz de Tenerife. El vie, 4 sept 2026 a las 9:17, Daniil ***@***.***>) escribió:

…

*danusha2345* left a comment (colbymchenry/codegraph#1641) <#1641 (comment)> Took this into a local integration build (current main + the open fixes I could merge cleanly) and ran the full suite with the native kernel: 245 files, 4323 passed, 0 failed. No regression on the codegraph repo's own index after a forced re-index. Looks good to me. — Reply to this email directly, view it on GitHub <#1641?email_source=notifications&email_token=ALJEAIYRD6GAXF4DYU5BHOD5NJ3ABA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNJTG43TCMRQGM4KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-5537712038>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALJEAI4GZHHZHX4APQBQDXL5NJ3ABAVCNFSNUABGKJSXA33TNF2G64TZHMYTCMZXGA3TQMRVGU5US43TOVSTWNJSG43TANBSGA2DLILWAI> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/ALJEAIYIO6J54GD35GCWRC35NJ3ABA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNJTG43TCMRQGM4KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG> and Android <https://github.com/notifications/mobile/android/ALJEAI5WSJIXZVOQCUWAS3D5NJ3ABA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNJTG43TCMRQGM4KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***> [ { ***@***.***": "http://schema.org", ***@***.***": "EmailMessage", "potentialAction": { ***@***.***": "ViewAction", "target": " #1641?email_source=notifications\u0026email_token=ALJEAI75TPFALJ6SVWUR7VL5NJ3ABA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNJTG43TCMRQGM4KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KWZ3NMFUWYX3DNRUWG2Y#issuecomment-5537712038", "url": " #1641?email_source=notifications\u0026email_token=ALJEAI75TPFALJ6SVWUR7VL5NJ3ABA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNJTG43TCMRQGM4KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KWZ3NMFUWYX3DNRUWG2Y#issuecomment-5537712038", "name": "View Pull Request" }, "description": "View this Pull Request on GitHub", "publisher": { ***@***.***": "Organization", "name": "GitHub", "url": " https://github.com" } } ]

[danusha2345]

Approving as a fellow contributor, with one caveat worth stating plainly: my review does not unblock the merge. I have no write access to this repository, so REVIEW_REQUIRED can only be cleared by @colbymchenry — this shows up as an approval in the list, nothing more. Treating it as a validation record rather than a gate.

What it rests on: the branch merged clean into a local integration build of current main plus the other open fixes, full suite with the native kernel 245 files / 4323 passed / 0 failed, and a forced re-index of codegraph's own repo with no edge regressions.

No rebase needed — GitHub still reports the PR MERGEABLE against today's main (unchanged since b9ca4b7, 31 Aug), and the branch is one commit ahead of the merge-base. Rebasing would only change the SHA.