A cycle start with no statement number is still a cycle #718
Loading…
Reference in a new issue
No description provided.
Delete branch "l2f-cycle-prologue"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
A cycle start with no statement number is still a cycle
The compiler gave the cycle start of the main procedure of SC-ILE/TFOR-STRIPPED
and of both SC-ILE/SCALE-AMBIGUITY twins no statement number: the entry code,
the cycle start and its compiler routines are one run of operations before the
first numbered statement.
statementOffound nothing for the jump back to thestart of the detail calculations, so that jump stayed
block_owner, and thecompiler routines after the cycle end stayed
remaining_effects.rpg_cycle.zig:prologueOfis the run before the first numberedstatement;
totalsuses it whenstatementOffinds nothing, so the cycleend is proved.
returnsBareis the final return (labels, a jump overlabels,
26) without the copy of X onto itself, which the compiler alsodrops when it drops statement 1000002.
rpg_cycle_template.zig:unnumberedPrologueproves the cycle starttemplate over that run, from the first
27 A; 37 45 72 B.rpg_specs.ziganalyze takes it when the numbered forms do not apply (one branch).
rpg_cycle_end.zig: the suffix of the body that no routine proved is thebare return when every jump into its labels comes from the entry code, the
cycle start or the body.
Public result: TFOR-STRIPPED statement 1500 and SCALE-AMBIGUITY statements 600
(both twins) recover; the SCALE statement 500 stays on
decimal_positions.twins-diff: no recovered group fell. emit-diff: no row worse, three rows
clean (failures 385 to 382). No effect on the held-out control_flow line
(oracle: unchanged). Tests:
src/rpg_cycle_prologue_test.zig.Not done, named next target: operation 33 (push the address of a symbol) fails
in the cycle-start statements of r7h-prt-ofl, r7h-prt-ho, r7h-prt-ind,
r7h-prt-one, r7h-prt-two and ACCESS (
110 111; 27 K; 37; 33 X; 35 T; 18, anaddress stored into a temporary; arm 33 requires a data symbol and X is not
one). FLOW/lx-total (SETON LR fused with the total calculation) is not done;
its ADD statements stop on
decimal_positions.Edit outside the lane files (src/core/rpg_source_unused.zig, L4, merged and idle; coordinator agreed): only the guard below. With the cycle body proved, the static initialization stores the zero default of
an unused packed field, and
rpg_source_unused.provedtook that value as proofof a source declaration with an unknown scale, so scale-one lost its
sparedeclaration. A zero packed default is no proof now (
defaultPacked).Gates