@alanc yes. (Disappearing now to pick up daughter...)

[illumos-gate] 15199 libdladm: unused label -- Toomas Soome <tsoome⊙mc>

[illumos-gate] 15200 libdladm: error: '&&' within '||' -- Toomas Soome <tsoome⊙mc>

so propolis is rust right? does that mean all the userspace emulation has the same

protection rust offers?

or is that mostly unsafe code so it doesnt?

OK that was a dumb question, i can just look at the code,

ok thats a lot of code, will have to look when i am not on my

phone

sjorge: isn't the interesting code in bhyve the actual device emulation?

sjorge: so does this actually do anything?

So I discovered a teeny-tiny test bug. If your login (say on a serial line) sets COLUMNS in the environment, the sed test suite will fail in one-out-of-130 tests, causing the suite to report failure.

I'm wondering if it's an actual test bug, or a "you should know better" situation.

sjorge: yeah, propolis is in rust, and mostly safe rust at that

nbjoerg: most of the device emulation is done in userspace

[illumos-gate] 15211 smbtorture failures in smb2.delete-on-close-perms.READONLY -- Gordon Ross <gwr⊙rc>

danmcd: In general I feel we should fix the tests to be as reliable as we can.

So we should look at how to fix that.

[illumos-gate] 15258 libsmb: tautological-pointer-compare -- Toomas Soome <tsoome⊙mc>

[illumos-gate] 15253 14993 broke test-runner -l option -- Toomas Soome <tsoome⊙mc>

rmustacc: Thank you for cashing the reality check.

jinni illumos 15266

jinni illumos#15266

→ BUG 15266: sed test is sensitive to COLUMNS in the environment (New)

→ CODE REVIEW 2564: 15266 sed test is sensitive to COLUMNS in the environment (NEW) | illumos.org/issues/15266

I never knew sed looked at COLUMNS.. there are some other conditions in there

we should probably just make the test always pipe the output of sed through cat - that would make it consistent regardless of COLUMNS or what TIOCGWINSZ returns

andyf: I tried changing the one failed test...

- $SED -n l $lines3

+ cat $lines3 | $SED -n l | cat

and it still fails with COLUMNS set.

Ah yes, the test is ' if (outfile != stdout)'

fair enough

The "l" function must always read COLUMNS and honor it.

Did update the comment to yours, however.

Thanks. I'm just wary of putting bug tracker links in source generally.

Ack.

If anyone from FreeBSD-land happens to be here, please ping me as I'm curious if my above bug affects FreeBSD or not?

danmcd: sed has some logic in it that is tty specific

I know.

and if it can't read the window size, it falls back to env

so that seems sensible

Which in the middle of a test script it can't.

test should sanitize the environment

nbjoerg: See my Gerrit link above.

Oh... NVM, you SAW it already.

right, change looks correct

nbjoerg: You can add a +1 on the gerrit if you deem it appropriate.