Make both boolean assertions negate the same way

[korya]

Problem

Two boolean assertions of the same shape did two different things with =false, and one of them told the user they had made no assertion.

$ http-assert --assert-ok=false http://…/500
[+] PASSED                                       ← asserted the status IS an error

$ http-assert --assert-body-empty=false http://…/ok
Error: Cannot perform request: no assertions defined

The second is the one that matters. The user named an assertion and was told there were none — the run cannot even be said to have failed, it never started. Nothing in the README, --help or the package comment described either behaviour, so neither was discoverable except by trying it.

Solution

=false now selects the opposite assertion on both flags.

$ http-assert --assert-body-empty=false http://…/ok
[+] PASSED                                       ← asserted the body is not empty

That reading is the one --assert-ok already had; it is the only one under which naming a flag can never yield nothing; and it adds a genuinely useful assertion by the shortest available route.

Invocation	Before	After
--assert-ok=false	asserts status IS an error	unchanged
--assert-body-empty=false	registers nothing	asserts the body is not empty

The pairing lives in one place now

The two flags drifted apart because nothing connected them — one was written with Changed(), the other with the value, and no rule said they were the same kind of thing. That is the same root cause as #35, where the fix was likewise to derive the rule rather than repeat it per flag: rejectRepeats reads the flag's type instead of consulting a list.

boolAssertion(cmd, name, whenTrue, whenFalse) is the equivalent here. A boolean assertion added later gets the behaviour on the day it is added, rather than whenever someone notices.

Deliberately not done

The issue floats promoting this to first-class --assert-not-ok and --assert-body-not-empty flags. That is a feature, not the fix — =false already works, is now consistent, and is now documented. Adding two flags for a spelling that exists would need its own argument about which is canonical, and it does not have to be made to close this.

Other Changes

Negation is documented in three places that previously described neither behaviour: the README's assertion section (with both flags shown doing something a reader would actually want), the --help prose, the two flag descriptions, and the package comment.

TestKnownIssue32NegatedBooleanFlagsDiffer is deleted, replaced by TestE2EAssertBooleanNegation — both flags, both directions, both outcomes, eight cases. Plus a guard that no negated flag is ever treated as no assertion at all, which is the specific failure this closes, and a check that repeating a negated flag is still refused, since naming it twice is still naming it twice.

Test_AssertBodyNotEmpty covers the new constructor including a nil body (what a 204 produces) and whitespace, which counts as content. It also asserts the pair are exact inverses across five bodies — without that, =false could quietly mean something subtly other than "not that".

There are no screenshots because there is no rendered UI; this tool's user-visible surface is terminal output, shown inline above.

Closes #32

Related:

• Let an empty body satisfy a body assertion #79 — merged; this PR was rebased off it onto master
• --assert-status overwrites while --assert-header accumulates, undocumented #35 — same root cause, fixed the same way: derive the rule rather than repeat it

🤖 Generated with Claude Code

https://claude.ai/code/session_019EMMhgmTkbzAsmeNy97PrP