should we perhaps ship usr/src/cmd/mdb/common/modules/conf/mapfile-extern ?

(or package I guess)

(to help any unbundled mdb modules)

Heh heh. fenix illumos#15655 breaks the smartos build as we set MAKEFLAGS in our smartos-live generated env file, but until this fix it was ignored by bldenv & nightly.

BUG 15655: Build scripts should tolerate make -w (Closed)

↳ illumos.org/issues/15655 | code.illumos.org/c/illumos-gate/+/2851

(Lucky us, easy fix is to add 'e' to our generated envfile's MAKEFLAGS.)

Slight fallout from illumos 15655 (fenix)

BUG 15655: Build scripts should tolerate make -w (Closed)

↳ illumos.org/issues/15655 | code.illumos.org/c/illumos-gate/+/2851

→ BUG 15663: bldenv fails with shipped .env files after 15655 (New)

andyf / danmcd: I'll get a fix for this out for review today

Oh wow, not just me then. Okay, thanks @sommerfeld for jumping back on this. (I thought it was a feature not a bug.)

hm, can any advocate look on 15588 pretty please?:)

tsoome: looking now

thanks:)

I think there was a period where for some bizarre reason I was not receiving RTI emails (checked my end for spam-filter failures and nothing... no idea). 15588 and at least one other (which I approved after reading the archives) were in that window.

no problem at all:)

I need to poke richlowe too, I guess:D

tsoome/gwr: Did anyone do ::findleaks measurements while testing the 15588 bits?

(if so, could you toss a note about it on the ticket)

not strictly necessary, I don't think, but would be a nice feel-good addition

Thanks sommerfeld, I think adding 'e' as you've said is definitely the best approach.

I concur Re: `bldenv will append an e if one is not present in MAKEFLAGS`

notably this will make a developer-visible change that predates 2005 finally take effect under bldenv (putting k in MAKEFLAGS, from the setting in the env file which was previously overridden by bldenv)

That's not great.

I think bldenv wants to still not have -k there if it can.

Otherwise it's always building with -m serial to figure out what's going on.

But fixing this is more important.

→ CODE REVIEW 2857: 15663 bldenv fails with shipped .env files after 15655 (NEW) | illumos.org/issues/15663

It really depends on what you're using bldenv for ...

I mean, I use it for trying to develop and figure out why I screwed up and the dmake -m serial in unix / genunix is already painful sometimes, but that's not important compared to the build confusion I think a couple of us hit.

Regardless, thanks for turning around quickly on this.

No problem, and as I said, sorry for the mess.

I'll be back in about 45 minutes.

45mins will be more than enough time to see if this fix alone cures what's ailing SmartOS. (And if it doesn't I'm ready to better-generate the env file in smartos-live.)

[illumos-gate] 15588 Improve efficiency of smb_ofile_lookup_by_fid -- Gordon Ross <gwr⊙rc>

rmustacc: I think the right thing to do for -k is to move it from the env files into nightly.

It may make sense in bldenv too. I realize that's just my gut reaction going, so may not be worth judging by that.

So there are two distinct use cases - "batch" and "interactive" compiles. Batch clearly wants -k; Interactive often doesn't (when I'm doing whack-a-mole fixes I generally run with -k first followed by a series of builds w/o -k)

I note that our make has a -S flag which undoes -k

(and gmake has that as well)

so interactive "make -S" when you don't want to keep going.

(hadn't noticed that flag before)

so if you can retrain your fingers to type "make -S" we can leave -k in MAKEFLAGS

As long as we are making a purposeful decision, I will adapt.

(and, oddly, poking through make's "main.cc": it looks like there's another undocumented way to do it: make -~k)

-S appears to be in POSIX, and GNU make accepts it.

wow, not even +k?

make people, why

(look for "invert_this" in usr/src/cmd/make/bin/main.cc)

I'll summarize the discussion here when I send 15663 for RTI, and capture it in the bug.

Thank you.

Just sent RTI 15663

Also, I sent an RTI for 15659 yesterday that I haven't seen any response on yet.

I was just telling people that I could approve them, but don't have the ability to run a test build of my own right now.

so if someone gets back to me on their feelings about that I can do the -Wformat one

I was hoping to leave the bldenv one to someone who had views on the chat above.

fenix illumos#15659

BUG 15659: libc genassym needs -Wformat fixes (In Progress)

↳ illumos.org/issues/15659 | code.illumos.org/c/illumos-gate/+/2852

I approved that, and robert I think was going to deal with it because of my temperature issues

Doing a local build for that now @sommerfeld. As for bldenv one, I'm a reviewer & tester so I'd prefer someone not-me to RTI-approve that.

fenix illumos#15663

BUG 15663: bldenv fails with shipped .env files after 15655 (In Progress)

↳ illumos.org/issues/15663 | code.illumos.org/c/illumos-gate/+/2857

the other I was leaving alone because I didn't keep track of the bldenv opinions, and I don't myself have one

[illumos-gate] 15659 libc genassym needs -Wformat fixes -- Bill Sommerfeld <sommerfeld⊙ame>

[illumos-gate] 15663 bldenv fails with shipped .env files after 15655 -- Bill Sommerfeld <sommerfeld⊙ame>

Draft heads-up in gist.github.com/Bill-Sommerfeld/e8b95c0600255e833a0fd45d19b94878