QA testing: how to run the file operation and shell cases
Two sections with the commands, a table of what each directory presses, and
an excerpt of a cases.tsv: how a case picks a target other than the panel
opposite, why most rows read the dialog instead of the result, that the files
are built again after a case moved them, and that the full filesystem is a
64k tmpfs at /small.
For the shells: what -s does, the list ci.sh walks, the six cases and what
each proves, and why the panel follows a cd under six shells and not under
mksh and tcsh.
QA testing: what the sandbox does not cover
A green run says nothing about another TERM, about spell with or without an
engine, about s3, mongo, docker, git and k8s, about a file that goes away
while it is copied, about a file root is not allowed to read, about FreeBSD,
or about colour and sixel, which a captured screen carries none of. The list
says so, so that a person knows what is still theirs to press.
QA testing: ci.sh, the CI job, the fileops and shells subjects
The page now names every subject there is, the shell axis (-s), the F6 and
F8 keys, the text:, no text:, clipfile: and no process left expectations,
and the sixth column of cases.tsv that names the shells a row is for.
A new section covers ci.sh, which runs every subject in one go, and the
ci-sandbox workflow that runs it on a push and on a pull request, with the
protocols left to the nightly run.
QA testing: the lua subject