CARD LAB3  written 2026-09-26 01:33:53 -0700 (clock-read; BEFORE the lab)
question: does January's pooling of a record prove that NO linked TU had an initialised definition of it?
lab: d.c = 'int a_one = 0; int a_two[5] = {0};' linked with a.obj (COMMON a_one 4, a_two 40), b, c, m.
predictions:
  R1 a_one is allocated in d.obj's own .bss contribution; the map names d.obj, not <common>; no pool entry.
  R2 a COMMON larger than the real definition (a_two 40 vs 20) draws a linker diagnostic (warning or error)
     rather than silently pooling.
consequence if R1 holds: a record January pools had ONLY tentative definitions in every linked object (so a
  tree that initialises it - e.g. actors.c '= NULL' - contradicts January), but pooling does NOT say which
  TU(s) held them.
