MONITOR blocks of ordinary statements recover #705
Loading…
Reference in a new issue
No description provided.
Delete branch "l2b-control-flow"
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?
MONITOR blocks of ordinary statements recover
A MONITOR whose body and handler are ordinary statements is proven by its
measured registration (
monitor.block, rpg_monitor.zig), not by the oneinteger conversion of
monitor.recover. The proof needs the 142 operationentry, a body that ends in the counter decrement, the jump to the end and the
handler label, the reset statement of ON-ERROR, a handler, the reset statement
of ENDMON, closed labels, and the measured descriptor, context and counter
links.
monitorProofs(rpg_shape.zig) builds the proofs of a procedure;shapeOfcuts the counter decrement and the jump from the body, so the bodyreads as source;
controlStatusno longer reportsmonitor_boundarieson thethree edges of a proven block (
monitorEdge).Not covered: nested blocks, several ON-ERROR clauses, a RETURN in the body
(MONITOR/limits keep
monitor_boundaries,block_owner,closing_destination).rpg_eval.zig
stepis unchanged. rpg_statements.zig net growth is -4 lines.Measured on the public twins (host build, Debug) against v0.1.415: 31 groups
turn recovered (decimal-fault-report, decimal-sign-report, packed-add-report,
packed-compose-report, packed-compose-sign-report, shared-io-full,
shared-io-none, DSPLY/error), 0 regressions; emit-diff: worse [], 8 better.
The statements of those programs that still fail (variable %SUBST index range
checks, WRITE effects) are other causes.
Gates