Skip to content

Unqualified MON 50 OPEN: does SINTRAN search current-user only, or fall back to (SYSTEM)? - CARVE (L07, segment 006-S3FS)

Answer for the nd500x / mon_path.c author. Question relayed:

On an unqualified 50B OPEN, does the resolver (GFILI/GCFIL) search only the current user, or fall back current-user -> (SYSTEM), or use a search-list / default-dir-index / friend-gate - and is the fallback SINTRAN's own or the linker's (LOAD emitting an explicit (SYSTEM) prefix)?

Source of truth: 006-S3FS.asm (this folder), whole-segment nd100-dis. Addresses octal; every citation is ADDR (line N) into that file. Version L-VSX-500 (L07). Symbols: 006-S3FS.symbols.txt (FILSYS-SYMBOLS). Binary: ../../../segments/006-S3FS.bin, load base 26000B, big-endian; byte offset of address A = (A - 26000B) * 2.

Evidence tags: [V] = I read the bytes at the cited addr/line and they say this. [I] = inferred from control-flow shape + symbol names, NOT fully byte-decoded. [OPEN] = not determined.

0. VERDICT (one line)

Current-user directory FIRST, then a SINTRAN-INTERNAL fallback to user (SYSTEM) - and the fallback is SUPPRESSED when the spec names a user explicitly. It is not "current user only"; it is not the linker's job. The whole thing lives inside GFILI in segment 006-S3FS. A caller does not have to build a "(SYSTEM)name" string to get the SYSTEM directory searched - SINTRAN does it.

Mapping to the request's options: it is (C)-then-(B) - a default-directory + directory-walk that terminates in an automatic (SYSTEM) scan, all internal to the file system. It is NOT (A).

This also matches the SINTRAN III Users Guide verbatim (independent confirmation, see section 5).

1. What runs on an unqualified OPEN

MON 50 OPEN worker OPENF @123525 calls the resolver once (FOPEN @067432 -> GFILI) and branches on the returned status; OPENF itself does no (SYSTEM) prefixing and no retry loop - the fallback is deeper, in GFILI. [V] OPENF entry 123525 (32097); single FOPEN call 123565 JPL I 46 -> ptr 123633=067432 (32129); no (SYSTEM)/retry in its body.

GFILI @057173 is the resolver (spec -> object entry). Structure [V]: - 057173 (13233) entry (SPUSH frame). - 057203 JPL I 152 -> ptr 057355=056645 = SEPOB (13241) - split the spec into subparts. - 057242 JPL I 121 -> ptr 057363=042622 = SEPFS (13272) - field split. - 057263 JPL I 102 -> ptr 057365=056326 = GOBJI (13289) - first object/file-name scan.

2. The default directory for an EMPTY directory/user subpart [V that GDEFD is called; I for gating]

When the spec has no directory prefix, SEPOB fills the directory from the caller's default directory via GDEFD @055263 ("get default directory"): - [V] SEPOB calls GDEFD: 057146 JPL I 20 -> ptr 057166=055263 (13211 / ptr 13227). - [V] GDEFD @055263 (12255) walks a per-user directory table: entry stride +030B (055364 AAA 30 (12320)), loop 055371 JMP -62 -> 055307 (12325), per-entry visibility-flag tests 055311 BSKP ONE 140 DA (12277) / 055313 BSKP ZRO 130 DA (12279), and returns a single directory index in ,B1/,B2 (055345-055351 (12305-12309)). - [I] GDEFD returns the user's ONE default directory; it does not itself auto-select SYSTEM. The exact "which flag = default" semantics are not decoded ([OPEN]).

So an unqualified name starts life pointed at the caller's own default directory - not a hardcoded index, and not SYSTEM.

3. The fallback: directory walk then automatic (SYSTEM) [V - decisive]

GOBJI returns via the SINTRAN skip-count convention (0-way = fail). Its fail (+0) return lands on 057264, which jumps into GFILI's tail block at 057405: - [V] 057264 JMP I 102 -> ptr 057366 = 057405 (13290 / ptr 13356).

Tail block 057405-057504 - the fallback engine [V]: 

057405  STA ,B2                         ; save GOBJI status                       (13371)
057406  SAT 56 ; 057407 SKP IF DA UEQ ST; status == 56 "No such file name"?       (13372-13373)
057410  JMP 4 -> 057414                 ;  yes -> try fallback                     (13374)
057411  SAT 57 ; 057412 SKP IF DA EQL ST; status == 57 "Ambiguous"?               (13375-13376)
057413  JMP 43 -> 057456                ;  neither 56 nor 57 -> return error       (13377)
057414  LDA ,B40 ; 057415 SAT -1 ; 057416 SKP IF DA EQL ST   ; ,B40 == -1 ?        (13378-13380)
057417  JMP 37 -> 057456                ;  ,B40 != -1 (user was named) -> NO fallback, return (13381)
        (,B40 == -1, i.e. UNQUALIFIED:)
057420  LDT ,B43 ; 057421 LDX 36
057422  JPL I 36 -> ptr 057460=050124 = GDIRT   ; advance to next directory        (13384)
057423  JMP -25 -> 057376               ;  more dirs -> loop back, re-scan GOBJI    (13385)
057424  JPL I 35 -> ptr 057461=010506   ;  GDIRT +1 return (list-end helper)        (13386)
057425  JPL I 35 -> ptr 057462=055540 = GSYSI   ; GDIRT +2: get SYSTEM user index   (13387)
        ... (057427-057442 set up SYSTEM dir) ...
057443  JPL I 23 -> ptr 057466=056326 = GOBJI   ; SECOND scan, under SYSTEM         (13401)

Byte-verified facts: - [V] GFILI calls GOBJI twice: first 057263 (13289), again 057443 -> 057466=056326 (13401 / ptr 13420). - [V] The fallback is entered ONLY when the first scan returns 56/57 (057406-057413, 13372-13377) AND ,B40 == -1 (057414-057417, 13378-13381). If a user was named, ,B40 != -1 and control returns the error at 057456 with no fallback. - [V] Between the two scans it advances a directory table via GDIRT @050124 (057422, 057433) and, when that walk is exhausted, obtains the SYSTEM index via GSYSI @055540 (057425, 13387). - [V] Symbols confirmed present: GDIRT=050124, GSYSI=055540, GMUSI=054527, GMAIN=047653.

Inferred (structure certain, exact semantics not fully decoded): - [I] The GDIRT loop walks the caller's directory search set; GSYSI is the terminal SYSTEM fallback. GSYSI = "Get SYStem Index"; it resolves SYSTEM dynamically by name (GSYSI -> GMUSI @054527, a name->index resolver), NOT from a hardcoded numeric user index. - [I] ,B40 == -1 == "no user/directory was named" (fallback enabled). Set from the parse in SEPOB/GFILI entry (057200 STT ,B40 (13238), 057217 STZ ,B40 (13253)); the exact predicate that produces -1 vs a real value is [OPEN]. The Users Guide (section 5) independently confirms the resulting behaviour.

4. The four sub-questions, answered

# Question Answer
a Current-user only, or fall back to (SYSTEM)? Falls back. Own default directory first, then user SYSTEM. [V] two GOBJI scans + GSYSI in GFILI.
b current-user -> (SYSTEM), or a search-list / default-dir-index / friend-gate? Both flavours: a default-directory-index (GDEFD) + a directory walk (GDIRT) that TERMINATES in a (SYSTEM) scan (GSYSI). Not a bare two-shot. [V] structure.
c Is the fallback SINTRAN's own, or the linker's (LOAD prepends "(SYSTEM)")? SINTRAN's own - it lives in GFILI @057173 inside 006-S3FS. The linker does NOT need to prepend (SYSTEM); OPEN does it. [V] GSYSI reached from GFILI's own tail block.
d Does the friend/public access check steer name resolution? No. GFIAC @057771 (ring/friend/public access) is not referenced anywhere inside GFILI's body - it is a POST-resolution permission gate run by the OPEN worker AFTER the entry is found. It does not choose which user/dir is searched. [I] (subagent grep of ptr-value 057771: hits only in the open/rename/set-access workers, none in 057173-057504).

5. Independent confirmation - SINTRAN III Users Guide

Reference-Manuals/ND-60.050.06 SINTRAN III Users Guide.md (lines 1720-1724) documents exactly this, for the same GFILI resolver used by @-command / program lookup:

"The search in the file-directories will be performed as follows: - The file-directory, where the user giving the command is given space, will be searched first. If not found, the user SYSTEM's file-directory is searched. - If a user name is specified in the file name, only that user's file-directory will be searched."

  • Bullet 1 = the GDIRT walk + GSYSI fallback (section 3).
  • Bullet 2 = the ,B40 != -1 suppression at 057417 (a named user disables the fallback).

The manual states the rule; the carve proves where it is (GFILI, 006-S3FS) and that it is internal to the file system, not the loader.

6. For mon_path.c (actionable)

  1. On an unqualified name (no (user), no directory prefix): search the caller's own default directory first, and on "no such file" fall back to the (SYSTEM) directory - implement this inside the OPEN resolver. Do not assume the caller/linker prepends (SYSTEM); it does not.
  2. Suppress the SYSTEM fallback when the spec names a user explicitly ((USER)name) - then only that user's directory is searched.
  3. Trigger is the file-name-not-found code 56 (octal, "No such file name"); ambiguous is 57. The fallback runs on 56/57, not on other errors.
  4. SYSTEM is resolved by name (GSYSI -> GMUSI), not a fixed user number - but for an emulator a fixed "(SYSTEM)" directory is a faithful enough model of the observable behaviour.

7. Still OPEN (honest)

  • Exact predicate that sets ,B40 to -1 vs a real value in SEPOB (the "was a user/dir named" test) - behaviour is confirmed by the manual, the byte-level setter is not fully decoded.
  • Whether the GDIRT walk reaches friend/public directories in addition to own + SYSTEM (the GDEFD visibility-flag tests at 055311/055313 hint at per-entry visibility, not decoded).
  • The literal string GSYSI feeds GMUSI (word 055560=010017) - not decoded, so "SYSTEM" rests on the symbol name + the manual, not on a decoded ASCII constant.
  • The 010506/010500 exit-dispatch helpers (057461/057463) - commoncode, below this segment.

Companion (name-MATCHING inside a directory): FLPAR-MDEAB-FILENAME-RESOLVER-CARVE.md (this folder). Status-of-record: SINTRAN/CARVING-HANDOFF.md.