test: the object-storage loops name the case each iteration tests (#982, 2 of 8)

[OffgridwithJD]

Second of #982's eight files, and the one that lost the most records. objstore_module: six checks sharing three keys, now zero.

Measured against a control run on clean main, same suite, same box:

clean main    30 records   24 distinct keys   3 colliding   6 lost
this branch   30 records   30 distinct keys   0 colliding   0 lost

Both rc=0, 30 checks, 0 FAIL. The record count is unchanged, which is how you can see this renames rather than adds or removes.

The two loops show both sub-shapes of the same defect

This file is a better pattern-setter than I expected, because its two loops fail differently.

Loop one — the headline already interpolated the scheme and only the continuation was short:

check     "a s3 URL reports an object-storage error, not a missing file"
check_num "and does NOT report it as a missing file"        # identical three times

Loop two — neither name carried the metacharacter, so three iterations produced one key for each of two checks. The headline collided as well, which loop one's did not.

So #984's rule covers both, with one extension: the continuation carries the same discriminator its headline names — and where the headline names none either, it gains one.

The discriminators were already in scope

From the loop variable, by parameter expansion — no new data, no invented labels:

scheme="${url%%:*}"                 # s3, gs, https
meta="${pat#s3://bucket/a}"; meta="${meta%.parquet}"   # *, ?, [0-9]

giving, from the run:

PASS  a remote glob (*) is handled remotely (an object-storage error, not a local one)
PASS  and the * glob is NOT reported as a local filesystem miss
PASS  a remote glob ([0-9]) is handled remotely (an object-storage error, not a local one)
PASS  and the https URL is NOT reported as a missing file

This is the mechanical quarter of #982 — six of the twenty-four. The remaining eighteen are hand-written sites where the discriminator has to be chosen rather than read off a loop variable.

One small thing, called out so it is not mistaken for scope creep

Loop one's headline now reads its scheme from $scheme rather than from a cut subshell, because both names need it. The resulting check name is byte-identical (a s3 URL reports... either way), so no ledger row moves on account of it and no name changed that did not have to.

No ledger change at all

objstore_module is not one of the two suites the ledger covers — those are harness_selftest and native_join_runtime_filter. So its check names have no rows, and this needs no regeneration.

That is also the sharp end of why #982 matters for #432 rather than for today's census: twenty-one of the twenty-four collisions are in suites that become covered only when the 240 are seeded, and they would go missing at that moment with every number still reconciling.

Gate

objstore_module   rc=0   30 checks   0 FAIL   on both trees
docs_style        rc=0   9 checks    0 FAIL
shellcheck        0 findings on the changed file
ledger + budget   untouched

A note on this PR's own commit message

It had to be written to a file rather than passed with -m: an apostrophe in the inline argument closed the quote and the shell then tried to execute the glob examples. Third inline-quoting fault of my session, against a rule that is mine — write the script, do not inline it. Recording it because the examples that broke it are the same metacharacters the suite is testing, which is a pleasing way to be reminded.

Part of #982. Six of twenty-four down; six more files.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Uf6UoeBRZYLQZa4KxNiw8a

[jdatcmd]

Approved at 0f6e1ef7. Verified with the same derived-helper method as #992 (write-up there, including two under-reporting failures of my own).

objstore_module.sh   names 23 -> 23
                     literal 22 -> 19, interpolated 1 -> 4

Three names move from literal to interpolated — the loop fix exactly: the continuation now carries the scheme its headline already had in scope, and the variable was already there. Name count unchanged, so this renames and neither adds nor drops an assertion.

Worth recording that my static sweep reported 0 literal duplicates on main for this file, against your measured 3 colliding. Both true: the continuation is one source line inside a three-iteration loop. A static count under-reports loops exactly as it over-reports interpolation — which is the third distinct way I have had that method be wrong today, and the reason #983's guard must count over emitted records.