You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Siblings of #5120: CICS resource-definition sources (SIT/SRT/JCT/TCT/MCT, JCL-assembled tables) #5154
Siblings of #5120 (found in the review of PR #5153). Class: CICS resource-definition sources read into the resource map (CSD decks and assembled macro tables, merged with provenance).
Covered by PR #5153: DFHFCT, DFHPCT, DFHPPT (program, mapset, partition set), DFHDCT (incl. ATI trigger), DFHTST, and CSD+macro merge.
Gaps:
DFHSIT (system initialization table)
DFHSRT (system recovery table)
DFHJCT (journal control table)
DFHTCT (terminal control table)
DFHMCT (monitoring control table)
Macro tables assembled inside JCL (inline SYSIN to the assembler step)
Firm up the ASSUMED operands (DFHPCT operand list, PPT PARTITIONSET=/PARTSET=, SDSCI RECFORM=UNDEFINED) from the CICS/VS and CICS/ESA 3.x Resource Definition (Macro) manuals
CEDA/CEMT runtime state (likely out of scope for a static scan; decide or refuse by name)
Rules: IBM is the reference (cite the manual and page); what IBM leaves open is ASSUMED or REFUSED by name. Pattern: gitgalaxy/core/cics_macro_tables.py from PR #5153.
Review of PR #5198 (covers SIT, SRT, JCT, TCT TERMINAL/REMOTE/SYSTEM, MCT EMP, in-stream JCL SYSIN, and refuses CEDA/CEMT by name in X62). Additional members of the same class still open:
DFHJCT: firm up the ASSUMED operands (JFILEID=, JTYPE=) from an IBM manual (only secondary sources so far)
DFHSIT: confirm or refuse the ASSUMED SUFFIX= operand, and decide whether to expand the SIT parameters (table suffixes, GRPLIST) into rows
JCL assembler step whose SYSIN is a separate member: join that member's tables to the job
The existing box for the PCT/PPT/SDSCI ASSUMED operands stays open.
PR #5263 closed the SYSIN-member join boxes (#5218). The checklist above (DFHSIT/SRT/JCT/TCT/MCT, JCL-assembled tables, ASSUMED operands, CEDA/CEMT decision) is still open, so this stays open.
Claimed by promoter-bot host melek, lease renewed 2026-10-30T20:09Z. It lapses after 6 h without a renewal. To take it back, assign someone or remove the claim:melek label.
Activity
squid-protocol commentedon Oct 15, 2026
Review of PR #5198 (covers SIT, SRT, JCT, TCT TERMINAL/REMOTE/SYSTEM, MCT EMP, in-stream JCL SYSIN, and refuses CEDA/CEMT by name in X62). Additional members of the same class still open:
JFILEID=,JTYPE=) from an IBM manual (only secondary sources so far)SUFFIX=operand, and decide whether to expand the SIT parameters (table suffixes, GRPLIST) into rowsThe existing box for the PCT/PPT/SDSCI ASSUMED operands stays open.
dead_key_audit: allowlist the ID and RECOVER macro operand keywords (…
CICS macro tables: DFHSIT/SRT/JCT/TCT/MCT and JCL-assembled tables in…
squid-protocol commentedon Oct 15, 2026
Siblings: #5218. These are the gaps left after the SYSIN-member join in PR #5217.
CICS macro tables: DFHSIT SUFFIX confirmed, MCT RECORD and TCT LINE/S…
CICS macro tables: SYSIN member join status, concatenated SYSIN, DFHA…
squid-protocol commentedon Oct 16, 2026
PR #5263 closed the SYSIN-member join boxes (#5218). The checklist above (DFHSIT/SRT/JCT/TCT/MCT, JCL-assembled tables, ASSUMED operands, CEDA/CEMT decision) is still open, so this stays open.
Fix ruff PERF401, mypy var-annotated, and update horror tag to H-0063 (…
DFHJCT: populate dsname from DSNAME operand (#5154)
Bump estate-crucible pin to v0.5.0 (#5154)
CICS macro tables: HLASM syntax audit, DFHPLT/DSNCRCT, JCL join, era_…
squid-protocol commentedon Oct 30, 2026
Claimed by promoter-bot host
melek, lease renewed 2026-10-30T20:09Z. It lapses after 6 h without a renewal. To take it back, assign someone or remove theclaim:meleklabel.