[Bakhomious]

What

guard-http, the next part of the guard (decision 51, fact 35): internal/guard now reads the options of curl, wget and gh api, and watches them only when they send data.

• A request writes when it has a method other than GET and HEAD, or, with no method, data to send (-d, -F, --json, -T, wget's --post-data, gh's fields and --input, and the like).
• curl's --next (-:) starts a request of its own; one that writes before it keeps the call watched.
• wget's abbreviations (--meth=POST) and its -e settings (-e Post_Data=x) count, as wget reads them.
• The values of short options are skipped, by each tool's help, so curl -H -d doesn't count as data.
• gh api graphql always counts, since GraphQL posts its reads too.
• Fail closed: an option, a method or a wget setting that only the run decides counts as a write; so do curl's -G with data and gh api -X=GET.
• Known misses, in decision 51: a whole word of options that only the run decides (curl $OPTS), ~/.curlrc, -K, ~/.wgetrc, WGETRC, and curl's other protocols (FTP's -Q, telnet://).

Checks

• mise run check: lint 0 issues, internal/guard at 100%, total 99.2%.
• Probes: 137 of 140 mutants caught. The 3 that survive change nothing: curl's data and form renamed are still the start of other names that send data, and the last match in lookup gives the same part as the first.
• Smoke: 28 scripts of my own through guard.Commands, each with the expected verdict.
• Requests to a listener on 127.0.0.1 confirmed the facts: curl -X GET -d -XPOST sends a GET with a body, wget -O- --post-data=x a POST, curl --output-dir a GET; gh api sends GET for -X=GET and -XGET, POST for -f a=b and --input, and GET with a query for -X GET -f a=b.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud