Composable Memory Transactions

• p. 6 Figure 2, Value production

  newTVar

• p. 8 Figure 4, rule (NEW)

  E[newTVar M]; Θ, ∆ ⇒ E[return r]; Θ[r ↦ M], ∆[r ↦ M]

• p. 8 Figure 4, rule (NEW)

  newTVar M

Judge's note: Page 6 crop: Figure 2 lists 'newTVar' with no argument. Page 8, Figure 4: (NEW) pattern-matches E[newTVar M]. A bare function name could be read as a partially applied value, so this is a notational inconsistency, not a definite error.

Second check: another model agreed

• p. 7 Section 5.1, last paragraph

  looks into the left operand of >>=, catch, and orElse.

• p. 7 Figure 3, Evaluation contexts

  E ::= [·] | E >>= M | catch E M

• p. 7 Section 5.1, last paragraph

  looks into the left operand of >>=, catch, and orElse

• p. 7 Figure 3

  [·] | >>= M | catch M

Only the authors can say what is right here.

Judge's note: Page 7 image: Figure 3 shows only the >>= and catch forms for E. In Figure 4 (page 8), orElse is handled by the separate rules OR1–OR3, which run M1 to completion with ⇒*. The prose may describe the combined effect informally, but as written it does not match the grammar.

Second check: another model agreed

• p. 7 Section 5.2, description of (ATHROW)

  ∆ are added to the initial heap Θ.

• p. 8 Figure 4, rule (ATHROW)

  Θ ∪ ∆′

Judge's note: Page 7 prose against the page 8 Figure 4 (ATHROW) conclusion 'P[throw N]; Θ ∪ ∆′'. The prose uses ∆ loosely for the allocation record, but in the rule the quantity added is the primed ∆′. Minor notational mismatch.

Second check: another model agreed

• p. 9 Section 6.4, first paragraph

  which places two additional requirements on the STM. Firstly, proper nested trans-actions are needed

Judge's note: Pages 9–10: Section 6.4 has no 'Secondly' or equivalent. The paragraph on waiting on the union of TVars when both branches retry is presumably the second requirement, but it is not labelled as such.

Second check: another model agreed

• p. 9 Section 6.2, foo example

  { x <- readTVar v

• p. 9 Section 6.2, following text

  If foo saw memory at a moment at which x-y was less than zero

• p. 9 Section 6.1 (continued)

  reads first consult the log so that they see preceding writes from the same transaction

Only the authors can say what is right here.

Judge's note: Page 9 image. Section 6.1 (pages 8–9) says each log entry records the old value when the TVar is first accessed, and that reads consult the log first. Under that implementation the second read of v returns the logged value, so x-y would be 0. The example presumably intended two different TVars; only the authors can say.

Second check: another model agreed

• p. 13 Appendix A, second version of f

  ; catch (g item) (recover exn item) }

Judge's note: Checked p.13 and crop C020-a. In the body of f, exn is bound only as a parameter in the where-clause definition "recover exn item = do { unReadPort p item ; throw exn }". The catch handler is a function from the exception to an STM action, as the 'bad' example above shows ("catch (f p1) (\exn -> f p2)"). So "(recover exn item)" at the call site refers to a free variable. This is a slip in illustrative code and the intent is clear, so it is minor.

Second check: another model agreed

• p. 2 Section 2.1, 'Secondly' paragraph

  rely on a a programmer-supplied boolean guard

• p. 2 2.1, blocking paragraph

  those that do rely on a a programmer-supplied boolean guard

Judge's note: The page 2 crop shows 'a a' in the right column, in the 'Secondly, blocking is not composable' paragraph.

Second check: none (judge only)

• p. 4 Section 3.5, first code example

  atomic (do {

Judge's note: Page 4 image: the example ends at the foot of the right column, and page 5 continues with prose, so the code is not continued there. Brackets are unbalanced: two opening braces and one closing; one opening parenthesis and none closing. Other examples, e.g. 'atomic (do { getR r1 3; getR r2 7 })', are balanced.

Second check: none (judge only)

• p. 4 Section 3.4, first paragraph

  STM Haskell also lets us to compose

• p. 4 Section 3.4, first paragraph

  STM Haskell also lets us to compose transac-

Judge's note: Seen in the page 4 crop, Section 3.4, first paragraph.

Second check: none (judge only)

• p. 7 Section 5.2, end of the STM transitions paragraph

  administrative transitions into the STM world, just as t.

Only the authors can say what is right here.

Judge's note: The page 7 image prints 'just as t.' literally, so this is not an extraction artefact. The intended ending (presumably a reference to (ADMIN) for the I/O world) is missing; only the authors can supply it.

Second check: none (judge only)

• p. 8 Section 6.1, second paragraph

  identical in the case a TVar that has been read but not written

• p. 8 Section 6.1, second paragraph

  the case a TVar that has been read but not written

• p. 8 6.1

  in the case a TVar that has been read

Judge's note: Seen in the page 8 image, Section 6.1, second paragraph.

Second check: none (judge only)

• p. 10 Section 6.4

  wait on the union on the sets of TVars

• p. 10 Section 6.4, paragraph on both alternatives calling retry

  wait on the union on the sets of

Judge's note: Seen in the page 10 image, Section 6.4.

Second check: none (judge only)

• p. 10 Section 6.6, second paragraph

  simple implemention of unbounded channels

Judge's note: Seen in the page 10 image, Section 6.6, second paragraph.

Second check: none (judge only)

• p. 12 Appendix A, first sentence

  we have become convinvced

• p. 12 Appendix A

  convinvced

• p. 13 Appendix A

  seems inconsisent

• p. 13 Appendix A

  rasies an exception

Judge's note: Seen in crop C017-a: the image prints "convinvced". The other Appendix A typos this cluster mentions (inconsisent, rasies, insside, "To to", arument) are judged on their own in C021–C024.

Second check: none (judge only)

• p. 12 Reference [4]

  Batchelors thesis

Judge's note: Crop C018-a shows the reference reading "Batchelors thesis, University of Cambridge Computer Laboratory, 2000." This is a misspelling.

Second check: none (judge only)

• p. 12 Reference [18]

  MARTNEZ, J. F., AND TORRELLAS, J.

Only the authors can say what is right here.

Judge's note: Rests on outside knowledge: that the first author of 'Speculative synchronization' (ASPLOS-X) is José F. Martínez. Crop C019-a shows "MARTNEZ, J. F., AND TORRELLAS, J." in the printed reference, so the letter is missing on the page itself and is not a text-extraction loss. This looks like a dropped accented character.

Second check: none (judge only)

• p. 13 Appendix A

  seems inconsisent

• p. 13

  seems inconsisent.

Judge's note: Seen in crop C020-a and on full page 13: "This treatment of effects that precede an exception seems inconsisent."

Second check: none (judge only)

• p. 13 Appendix A

  if the computation rasies an exception

• p. 13 Appendix A

  the computation rasies an exception

Judge's note: Seen in crop C020-a and on full page 13: "should be reverted if the computation rasies an exception."

Second check: none (judge only)

• p. 13 Appendix A

  “look insside” a catch. To to this

• p. 13 Appendix A

  look insside” a catch. To to this

Judge's note: Full page 13 shows "one that does not “look insside” a catch. To to this, we add a new STM evaluation context".

Second check: none (judge only)

• p. 13 Appendix A, final paragraph

  run the first arument of catch

• p. 13 Appendix A, after XSTM rules

  the first arument of catch

Judge's note: Full page 13, the paragraph after the XSTM rules, shows "All three rules run the first arument of catch".

Second check: none (judge only)