Opened 7 hours ago
#444 new defect
ISOLTS hdump on an L68 prints wrong history-register data (pas2 overwrites its own DPS-8 test)
| Reported by: | Eric Swenson | Owned by: | Eric Swenson |
|---|---|---|---|
| Priority: | minor | Milestone: | MR12.10 |
| Component: | Tools | Version: | MR12.5 |
| Keywords: | Cc: |
Description
In the ISOLTS executive pas2 (pas.061, "multics pas executive rev. b.2 062287"), the L68 history-register print loop advances the wrong instruction. The code at L45330 (LDA 1887 / TZE L45333 / TRA L45500) is the DPS-8 test. The loop treats L45330 as the per-entry load: 045432–045434 add 2 to it each row, and 045444 resets it to LDA 46234. The real per-entry load, L45333 (the row's flag bits), is never advanced.
On an L68:
- the first hdump repeats entry 0's flag bits on every row and takes rows 2–16's addresses from pas2 words 1889–1917;
- every later hdump tests CU entry 0 instead of word 1887 and, once history has recorded anything, prints the DPS-8 format with 64 entries, 48 of them never read.
DPS8/M CPUs are not affected. Fix: refer to L45333 instead of L45330 in the instructions at 045432, 045434 and 045444 (three words in the deck).
Note:
See TracTickets
for help on using tickets.
