D096 - Two answers to the symbol count, and the divergence is kept rather than reconciled

Status: decided. 2026-09-09.

Info::symbol_count refuses to infer a count and dynamic::symbols infers one. Both are right, and what was wrong is that neither said so.

Orbistoun's symbol-level differential asked whether symbol_count returning None should be surfaced. Looking at why it can returned something better than an answer: the crate already contains two functions that answer the same question differently.

• Info::symbol_count returns symtabsz / syment, and None when either is zero. Its doc is explicit about why: "a symbol count inferred from a table's extent is a count that silently changes when something else moves".
• dynamic::symbols makes exactly that inference. With no symtabsz it walks to the end of the segment, so its length is bounded by the segment rather than by the table, and anything trailing the symbols is decoded as more of them.

Kept, not reconciled

This is the same shape as D091 - two functions, one crate, one question, opposite standards - and the resolution is different, because this time the two standards are both correct for where they sit.

D091's line was that a writer must refuse where a reader may guess. symbol_count is a claim about the file: it is asked "how many symbols does this module say it has", and the honest answer to a module that says nothing is None. symbols is a reader: one that stopped dead on a missing size field would read nothing at all where it could read almost everything, and a wrong line on a terminal is the worst its guess can produce.

So the fix is not to make one behave like the other. It is that a caller comparing them had no way to know which of the two answers it was holding, and now both docs say. A caller that needs the exact count asks symbol_count and treats None as this module does not state one.

Why it is not being changed on the evidence available

Latent, not observed: orbistoun's differential found symtabsz present on all 29 modules of its corpus, and this crate's own writer always emits it. Nothing has been seen to take the inferring path. Principle 5 - the change to make is the one material asks for, and no material has asked for one.

The one thing worth watching, recorded so it is not rediscovered: imports iterates symbols, so on a module without symtabsz it would decode whatever trails the symbol table as further symbols. They would almost all fail decode_symbol_name and be skipped, which is why this is a latent oddity rather than a live defect - but "almost all" is not "all", and a plausible eleven-character string in trailing data would enter an import list. If a module without symtabsz ever turns up, that is the thing to look at first.