Automate the subr_unit test.
ClosedPublic
Actions

Authored by asomers on Apr 20 2016, 11:37 PM.

Details

Reviewers

ngie
gibbs

Commits

rS298811: Automate the subr_unit test.

Summary

Build and install the subr_unit test program originally written by phk, and run
it with the other ATF test.

tests/sys/kern/Makefile
Build and install the subr_unit test as a plain test

sys/kern/subr_unit.c

	  command-line parser to override it.

Repository

rS FreeBSD src repository - subversion

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 3411
Build 3449: arc lint + arc unit

asomers retitled this revision from to Automate the subr_unit test..

asomers updated this object.

asomers edited the test plan for this revision. (Show Details)

asomers added reviewers: gibbs, ngie.

tests/sys/kern/Makefile
29	FYI, I'll be able to remove the -Wno-sign-compare after merging D6004

Should a similar approach be used with to tests/sys/kern/acct/acct_test here ?

tests/sys/kern/Makefile
7	Please use SRCTOP instead.
28	This should be fixed...

Use SRCTOP and justify disabling compiler warnings.

tests/sys/kern/Makefile
28	This is a weird one. There are a lot of functions, such as clean_unrhdrl, whose prototypes are in headers that cannot be included in userland. In the test program, though, they're effectively static. There are three options: Add prototypes in an `#ifndef _KERNEL` section near the top of subr_unit.c. I don't like this option because it isn't DRY. Move the relevant prototypes into their own header, subr_unit.h, and include that from systm.h. I don't like this option because it seems heavy-handed and there's no precedent for it. Add -Wno-missing-prototypes. If you agree that this is better than option 2, then I'll add some comments about why I did it.

Are you satisfied, ngie?

sys/kern/subr_unit.c
888	I've probably remove "rather than ... kern" because that can change at a later date.
943	I'd probably use a macro/put this up near the comment about why things are the way they are.
986–989	This is not properly sorted (char* -> long -> int -> u_int -> bool)
996–997	How do you know errno hasn't been set to a different value by another preceding call? The official way to do this (per opengroup) is to set errno = 0, then test out errno after this call: http://pubs.opengroup.org/onlinepubs/7908799/xsh/strtol.html
1019	a != NULL?