danmcd - that's an interesting side-effect of the isatty() change that feels like a new bug

a successful call to printf() should not set errno

Filed issue 15294 (fenix)

BUG 15294: Successful *printf() calls should not set errno (In Progress)

↳ illumos.org/issues/15294

hm. ::stack does not present proper number of function parameters:(

tsoome_: it should use ctf, but i think it's still capped at 6 if a function uses more

andyf: If your bug is correct/actual/etc., then 15291 should be marked as a subset of yours.

I couldn't reproduce it with a toy, FWIW. I suspect there needs to be long buffering or something.

And it might be strerror for all I know?

jbk ah, that explains, this one had 8

yeah, it's actually a reason why I try to avoid writing functions with more than 6 arguments... yeah you can fish out the arguments from the stack, but it's kinda annoying

danmcd - I've added a test as part of code.illumos.org/c/illumos-gate/+/2582

→ CODE REVIEW 2582: 15294 Successful *printf() calls should not set errno (NEW) | illumos.org/issues/15294

The ZFS tests still needed fixing, they shouldn't assume errno is unchanged over other function calls

If you ran your toy and piped it through cat, it would probably have reproduced. The output needs to be not-a-TTY

[illumos-gate] 15132 libsmb SMF initialization problem messages incorrect -- Andy Fiddaman <illumos⊙fn>

andyf: I'm just catching up, but did you check for other people who isatty and don't preserve errno?

Sonofabitch @andyf, you're right.

When I'm back (firefighting has kept me from proper time off, so I'm extending what I'd planned into most/all of next week) I can test this in lieu of fenix illumos 15291

BUG 15291: zfs-tests errno flaws exposed by 15220 (New)

↳ illumos.org/issues/15291 | code.illumos.org/c/illumos-gate/+/2579

(right now that only exists in the 20221229 branch of illumos-joyent).

Okay back to being off...

richlowe - rmustacc did a survey of isatty() as part of issue 15220 (fenix) - we were the odd one out in terms of always preserving errno, and now we're aligned with what people expect (POSIX technically allows either).

BUG 15220: isatty should return reasonable errnos (Closed)

↳ illumos.org/issues/15220 | code.illumos.org/c/illumos-gate/+/2540

If that's what you're asking

No, I'm asking whether there's other code like *rintf that maybe calls isatty inside itself, but doesn't expect errno

Ah, got it. I checked usr/src/lib/ and fixed up a couple of other callers in there as part of the same change.

I should probably add a second bug ID for that bit

I looked briefly through usr/src/cmd too, but not exhaustively.

[illumos-gate] 15295 SMB services should check IPC caller privileges -- Gordon Ross <gwr⊙rc>