Rest-of-the-story ==> stupid .env file mistakes that didn't get caught until the gcc10 switchover.

Ah, thanks danmcd - I regularly do RTI builds on r44 and builds for onu on bloody, I was just about to do a build using the env files under /opt/onbld. We do those as part of the pre-release process, but I probably have not done one since.

My build on r44 using /opt/onbld/env/omnios-illumos-gate was successful

(not yet updated for gcc10)

I had PRIMARY_CCS & PRIMARY_CCCS because of stupid M-w/C-y technique. Fixed those and BOOM it's all good.

does omnios have something similar? OpenIndiana/oi-userland #11418

We will be doing the same, no doubt.

hm, and how can one set default mediator...

mediator-priority=vendor

I think OmniOS does it for ctags, so I'll find a link

ah, ok

I see

I'd guess, we rather would like to default to tcsh:D

I think so, yes, and it would be nice if OI and OmniOS used the same implementation name

mediator-implementation=tcsh mediator-priority=vendor

yep

I think hadfl said that we'd have to wait for the new `which` to land first btw, since the old `which` script doesn't work properly if csh is really tcsh

right. so he better hurry:D

I think that's close - illumos 15471 (fenix)

FEATURE 15471: replace csh based which(1) with implementation in C from BSD (New)

↳ illumos.org/issues/15471 | code.illumos.org/c/illumos-gate/+/2710

There's just an interesting balance around backwards compat, and being consistent with tcsh's builtin and how GNU which behaves.

We apparently can't satisfy both

true.

and GNU which expects the output of `alias` to be fed into stdin for full operation

eh, should have patched which to use /usr/has/bin/csh in 15476.

I think 15471 has taken longer than suspected

but there's just the one query left I think, which is just whether the 'no <cmd> found in PATH' message goes to stdout or stderr

ye, nothing fatal, but for correctness :)

Oh yeah, which is builtin in tcsh.