bmake doesn't search for BSDmakefile, the way that gmake searches for GNUmakefile, right?

hu

GNUmakefile is nice for disambiguating the implementation language, in terms of helping text editors to hightlight, and linters to lint correctly

no, though you can set .MAKE.MAKEFILE_PREFERENCE to change that

github.com/neosmart/gmake-proxy seems to say it works

Title: GitHub - neosmart/gmake-proxy: A BSD make (bmake) proxy that invokes GNU make (gmake) instead

I've `mv Makefile BSDmakefile, yup, that works

well `git grep -i bsdmakefile contrib/bmake` gives no results

so not sure how that one's supposed to work

Hmm... I use GNUmakefile to hold GNU make specific syntax which enhances my build environment but is not required functionality.

That way I can sneak in things that help but won't break BSD make for example.

`find . -type f | xargs wc -l` gives 101 930 lines

will be fun to hunt that "BSD"

1044 lines contain "BSD"

What I've found in various parts of the source code is it accepts also "makefile"

Found it, jrtc27 it's the variable you said beofre

share/mk/sys.mk

317:MAKEFILE_PREFERENCE?= BSDmakefile makefile Makefile

but that's because of freebsd, not bmake

bmake itself doesn't do that

and netbsd's mk fragments presumably don't either

Author: Simon J. Gerraty <sjg⊙Fo> Date: Wed Apr 3 16:20:21 2013 +0000 Tell bmake to use the FreeBSD preferred makefile preference list. PR 177593

o bin/177593 sjg [PATCH] bmake(1) doesn't honor BSDmakefile like pmake

bugs.freebsd.org/bugzilla/show_bug.cgi?id=177593 doesn't comment about what happens in other BSD flavours

Title: 177593 – [PATCH] bmake(1) doesn't honor BSDmakefile like pmake does

by the way, speaking of fragments, I've seen some subdir/[mM]akefile in the code, so yeah, doesn't seem rocking solid

preyalone: can we help you to write a Makefile compatible with BSD and GNU Makefile perhaps?

oh, i'm already targeting the POSIX subset

i'm actually writing a linter to encourage extreme portability

for GNU extensions to the POSIX standard, the linter can recommend two courses of action:

either remove the GNU-ism, or else rename the file like GNUmakefile or *.GNUmakefile

so i would like for bmake to introduce a similar BSDmakefile search, which would process after makefile and Makefile

however, i am seeing strange behavior with bmake

Homebrew's bmake does not appear to run BSDmakefile

nor does bmake's man page mention BSDmakefile

Yes that's how FreeBSD make world comiles it this MAKEFILE_PREFERENCE optin

+ if you've BSDmakefile and Makefile in the same dir, BSDmakefile wins

oh drat! only *FreeBSD*'s implementation of bmake supports BSDmakefile??? not bmake for Homebrew, bmake for other BSD OS's?

the difference in load order is also awkward

i think gmake loads GNUmakefile last not first

Title: BXR.SU — Super User's BSD Cross Reference — Search

it seems FreeBSD specific

Title: GitHub - mcandre/unmake: a makefile linter

got a basic AST going for the POSIX subset

i'm most afraid of screwing up in the AST part of my linter. so if you have any makefiles, gnu, bsd, or otherwise, happy to hear how unmake behaves with them

did a quick test with 3 files, gmake takes also GNUmakefile first

GNU make order seems to be GNUmakefile makefile Makefile

can freebsd please update its man page for bmake to note BSDmakefile?

By the way, GNU make only added support for != in 4.0

(10 years ago yeah)

so be careful with the assertion GNU Makefile = POSIX Makefile + additions ; BSD Makefile = POSIX Makefile

lol

i mean, i could go to the effort of maintaining a table of implementation to features, similar to that one table of es6 browsers and features, or the table of c++23 compilers and features

frankly, there's enough basic linter rule fodder like "don't use del," (Windows specific), "don't use rm," (UNIX specific), etc. to last me awhile

i've been citing gmake v3 as a minimum requirement in my projects... forgot which feature v3 needed. but maybe i'll recommend v4 now

except i can't, because xcode is still stuck on gmake 3.81

i assume that most conditional logic involves an extension, regardless

oh it's an "execute and assign operator"

in GNU make 3 you needed to do FOO = $(some command)

BSD make doesn't support that syntax but supports FOO != some command

4 added support

execute and assign, hmm. i'm used to the GNU feature VAR=$(shell <command>), which behaves similarly to VAR="$(...)" string interpolated shell commands in sh family

i'm frequently amazed to learn which extremely useful features turn out to be extensions. i been documenting the hell out of my dev environment requirements, for basic and subtle needs like GNU findutils... curious if BSD findutils gets -print0 / xargs -0, and -o unioning subqueries together

i was wrong about gnu load order. gmake loads GNUmakefile first

preyalone, find(1) says -print0 and -or are both available.

:)

But that same manual page says -print0 isn't POSIX.

iirc -print0 is on the verge of POSIX inclusion

That would be nice.

(implied but not stated is that xargs -0 is as well, to complement it)

I don't recall where I heard that, though

oh, no, I think they're in: austingroupbugs.net/view.php?id=243

the "severity" field threw me off when I first opened it, but the status is "applied" and "accepted as marked"

find -print0 | xargs -0 came into existence before -exec ... {} + appeared. Once {} + appeared there was no more specific need for | xargs -0 anymore.

github.com/BSDNow/bsdnow.tv/blob/ma…2/feedback/Fred%20-%20Networking.md this sure has happend to me fighted with that at Cityterminalen before going to Prauge

Title: bsdnow.tv/Fred - Networking.md at master · BSDNow/bsdnow.tv · GitHub

anyone knows when 802.11ac will be available -_-~~~

hm, freebsdfoundation.org/project/wifi-update-intel-drivers-and-802-11ac only says in progress, no idea by who

Title: WiFi update – Intel drivers and 802.11ac | FreeBSD Foundation

oh, I should skim better "The Foundation is funding Bjoern Zeeb to …"

already updated to the next release?

my laptop at least updated just fine without issues

I've updated 12 servers fine

yeah

but nothing in the release notes was especially worrying from a server POV

i am just a desktop user but it went fine here as well

_0pr_: ah, yes there was freshports.org/net/wifibox to use a virtualized linux to use 802.11ac via that

Title: FreshPorts -- net/wifibox: Wireless card driver via virtualized Linux

minor versions .x usually go fine but then when there is major version change then there might be more issues

Cant remember who it was that had problems with connecting to a certain hotell wifi network

irc logs to the rescue =D

Gud - not here :-(

Title: bsdnow.tv/Fred - Networking.md at master · BSDNow/bsdnow.tv · GitHub

I don't understand how anyone can think that encryption can exist without either PKI or PSK.

I suppose in theory Alice and Bob can also agree to meet in a dark alley under the fourth harvest moon and perform some sort of eldrich ritual, but that's not really scalable.

we should just all exchange big OTPs with each other n^2 on a global scale seems like a good idea

Ah yes, one true pairings are sure to solve everything. :P

I meant one time pads, should have just written that in full

Too late, I like the idea of one true pairings better.

antranigv: how's the jail review going?

Oh, that thing that I never know what it's called is called GEOM whithering.

Are there any plans to port the "nouveau" graphics driver to FreeBSD in the near future?

GEOM whithering is what causes other instances of the same GEOM device to disappear when you open one.

msiism: that's probably a better question asked of the drm team in #freebsd-xorg on EFnet.

withering?

debdrup: Okay, I'll if I can make it there. Thanks.

github.com/freebsd/drm-kmod is where the work happens, so you can also check there.

Title: GitHub - freebsd/drm-kmod: drm driver for FreeBSD

Reinhilde: I... just explained it two minutes ago.

interesting.

g_wither_geom(9) has a bit more information.

geom(4) also has a subheader on ORGANIZATION

Err, the manual page quite clearly says ORPHANIZATION

any guides on tuning startup time?

hey folks, i finally got the kernel crash dump from my gpu related crashes. it has a full vmcore dump and a gdb trace at point of failure. the error i saw was something like 'fault at a nofault page'

*fault at nofault entry

stacktrace starts from the nvidia.ko driver

i've gather a bunch of info regarding this, i just popped in to ask in which category to open the bug report. Doesnt seem like the base system but i guess it would be the 'Other' category

I have 2xSSDs (500 GB each) created zfs stripe on both disks, and my website files are running on those disks. Later I added 6xSATA HDDs to my system (they will be hosting big video files within my website), in that case, do you recommend me to set a compression to my dataset for the videos, on SATA disks? If so, what kind of?

Thanks in advance.

tercaL: video is usually well compressed anyway, I thought. easily tested though - create the filesystem, copy in, check, then decide.

iirc, zstd and lz4 have a quick check when saving to check if compression helps and store uncompressed if it does not

Yes. on modern cpu you should not notice compression of incompressible data

now upgraded to 13.2-RELEASE

went without problems

(desktop system with Xfce4 etc)

LapsangS: I would hope that after six RC we caught enough bugs :D

yeah

well i thought that this was safe to install after first installing in on my old test-laptop

and it didn't explode or give any mysterious error messages

dch: So, I shouldn't even worry about enabling the compressing, or it would still help?

tercaL: it's enabled by default, meaning it does not hurt in any use case

when i create new pools I enable zstd regardless what I'm gonna put on it

I mean, usually it's SQLite databases…

i use zstd too, but because it's new and cool :D

yuripv: new and cool can stay far away from my data storage.

i want old, boring, consistent, mature when it comes to data storage

Demosthenex: I think compression, even the newest ones ZFS supports is probably at least 8 years old

if i install kitty terminal from ports, from my understanding, i have to live with all those python pkgs installed?

after i build, i mean

meena: you saw zstd just had a data corruption bug? facebook/zstd #3517

Title: [bug-fix] Fix rare corruption bug affecting the block splitter by terrelln · Pull Request #3517 · facebook/zstd · GitHub

made hackernews front page :P

what do you mean by 'all those python pkgs'? I only see python itself as runtime/library dependency there, no other python package

trev: pkg autoremove doesn't do it?

Demosthenex: I generally ignore hacker news…

meena: not sure yet cause the build is taking for.ev.er....

meena: better than reddit ;]

general tip, instead of building ports directly on your system build a pkg repo with poudriere or something and install from that pkg repo

Demosthenex: not sure i'd agree with that. reddit is a specifically different kind of bad

trev: what nimaje said

oh wow

didn't know that..new to freebsd

is poudriere the most widely used?

trev: pretty much, yes

are there any third-party repos that i can add that build ports regularly or something?

I think you just want to switch to the latest repo

i am on latest, just trying to avoid building myself

i will build it in poudriere though if that is best case scenario

latest is build regulary from ports, what benefit do you hope to get from a third-party repo?

a pre-built binary

(quarterly is too, but from a quarterly branch that gets much less changes (should be only security fixes, build fixes and metadata fixes afaik))

i guess i am misunderstanding something. there is a "latest" pkg repo where ports are prebuilt?

yes, latest is a pkg repo, that is build from the main ports branch (quarterly is build from a quarterly branch

create a /usr/local/etc/pkg/repos/FreeBSD.conf file containing FreeBSD: { url: "pkg+http://pkg.freebsd.org/${ABI}/latest" }

done, made the report regarding my gpu problems

nimaje: thanks a lot. just realized this is in the handbook and i glossed over it when reading about poudriere. next time tell me to RTFM :D

still don't see ports pre-built on latest

just latest versions of the in-tree binaries (or whatever you call it)

nevermind, it's just that the package i wanted to install is not pre-compiled

damn

hello folks, how can I display my loginClass? id does not limits either racct in not enable

damned, I need to learn how to read. Forgot my stupid question (of course id does, with -c flag).

Another satisfied customer. . .

:-)

Today I found sysutils/httm and it's pretty cool

When installing a port, you usually get a TUI window with package config options. How are the options edited (I would like to add --with-features=huge to the ./configure of vim)

ElectricJozin: you can't/shouldn't

`make showconfig` brings up dialog4ports again.

They _are_ stored on-disk, but editing it can be very fraught - and you get much better results by using make.conf(5) as that means you only have to make modifications one place.

No, but 'make config' does

Right.

Shows long long it's been since I started using make.conf(5) that I can't even remember that..

ElectricJozin: If you want to add an option that doesn't exist in the port, you'll have to edit the port's Makefile. But take care in doing so, there might be lions, tigers and bears there.

vkarlsen, well the configure option is added in the makefile's CONFIGURE_ARGS, I want to edit the tui to make it easier for people

and then push it as a change

ElectricJozin: Maybe you could add a flavor for it? I see tiny is a flavor already.

whats a port flavor?

It's an easy way of specifying an alternative package made from a port, if memory serves.

debdrup, your memory matches mine, but they left.

This is driving me nuts. Can't figure out why, for some reason, my rc script isn't able to run the process as a regular user. It uses command=/usr/sbin/daemon and command_args has -u $USERNAME (confirmed it's a valid user)...

But when it's run as -u $USERNAME, it fails. When that's removed and so it runs as root, it works. I'd prefer for it to run as a dedicated, unprivileged user though.

What's also confusing is ${name}_user (a la /etc/rc.subr). /usr/sbin/daemon, which would be the initial process, needs to run as root so that it can set limits (e.g. login class) & then invoke the main child process using -u $USERNAME.

So it would be that /usr/sbin/daemon is run as root:wheel and then the main child process it's starting/supervising is run as the specified user. But ${name}_user runs the entire command, supervisor process + child process, as that user, which fails due to it not being able to invoke su -m $USER or limits.

I wondered if the home dir and/or shell set for the unprivileged user would make a difference, but apparently not. Would prefer it to be /nonexistent and /usr/sbin/nologin anyway.

Someone want to take a quick look at /etc/rc.subr:1124 for me and tell if that looks like?

*right

Shouldn't -c be followed by a login class? It looks like it's expecting to pass a command string there?

...and that shouldn't be there to begin with since by default it starts with "limits -C ..." (where ... by default is 'daemon')

Unless I'm missing something.

1124 is "fi" for me

Mind termbin'ing what you have at that block for me? Surrounding 5 or 6 lines before or after should do.

it would be easier if you would pastebin what you are looking at :)

doit="su -m $_user -c 'sh -c \"$_doit\"'"

What do you see for that line? Should be right around 1124.

-c means something different depending on whether it's before or after the username in su

su -c class someuser -c command

oh good lord

I was wondering if it was something like that.

this is because everything after the username is actually a parameter to the shell, NOT to su

blame the gnuists for breaking the clear distinction between options and arguments

Reading su(1) right now.

I was reading it earlier, but I somehow missed the line that says that.

it's somewhat abominable that it has to use -c 'sh -c ...' with the resulting double layers of shell quoting

it doesn't, but it's in there in examples

Yeah, it's confusing.

honestly the only reason why it's not a security nightmare is that all the inputs are from supposedly trusted places

For whatever reason, I keep getting: `daemon: /usr/local/bin/consul: No such file or directory`

But it works fine when I run it as root, which is what I had been doing so far. I wanted to clean it up and run it as a dedicated user.

PATH issues?

I checked /usr/local/bin/consul, and owner/perms look fine

That's what I was wondering too, but PATH looks fine from what I saw. Plus it's finding /usr/sbin/daemon and all of the paths are absolute anyway.

check also if /usr/local/bin/consul is a script file and whether its interpreter is present

It's an executable.

133mb executable since it's Go-compiled and has embedded assets (web app).

what are the full args to daemon?

one sec

/usr/local/etc/rc.d/consul: DEBUG: run_rc_command: doit: limits -C daemon /usr/sbin/daemon -c -f -t consul -u consul -p /cluster/consul/pids/consul.agent.pid -P /cluster/consul/pids/consul.supervisor.pid -o /cluster/consul/logs/consul.log -H /usr/local/bin/consul agent -config-dir=/cluster/consul/conf -log-level=info -log-json

Everything exists and has proper ownership+perms too.

Though I did just notice something...

/usr/local/bin/consul: ELF 64-bit LSB executable, x86-64, version 1 (FreeBSD), statically linked, for FreeBSD 12.3, FreeBSD-style, Go BuildID=a7sO-eU5nsTRWwyjoGPA/jjd432ImPUDe9rvlcHXj/vDviKQ68exEgSX6-fFMv/qW7IyKOpu05SL6AfOjnx, with debug_info, not stripped

This is 13.1-STABLE, and this executable seems to be built for 12.3.

Would that cause an issue?

I'd expect a different error if it did

Complaint about ld-elf or something I take it?

presumably you have COMPAT_FREEBSD12 in the kernel

I'll double-check, but it should.

assuming it really is static...

Yeah, kern.conftxt shows COMPAT_FREEBSD12

Hmm, going to rebuild the source-compiled executable I had before and try that to make sure it's not a problem with the actual executable (though given that it works fine as root, I can't imagine so?)

what does readelf -dl consul output

ok, so definitely static

Can go generate dynamic binaries?

and perms on / /usr /usr/local /usr/local/bin /usr/local/bin/consul ?

Yes

Huh.

Does that by default if -static isn't passed and CGO_ENABLED=1, I think.

RhodiumToad: Let me confirm perms, we're looking for root:wheel & 755, right?

yup

Yup, all are root:wheel & rwxr-xr-x

Everything under /cluster/consul was 700 (and 600) before, but I changed those to 755 (and 644) for now for troubleshooting.

and it works if you run the daemon command yourself?

Yes, let me try again just to be sure. I'll paste the command that does work.

This works when I run it from a root shell: `limits -C daemon /usr/sbin/daemon -c -f -t consul -p /cluster/consul/pids/consul.agent.pid -P /cluster/consul/pids/consul.supervisor.pid -o /cluster/consul/logs/consul.log -H /usr/local/bin/consul agent -data-dir=/cluster/consul/data -config-dir=/cluster/consul/conf -log-level=info -log-json`

I left out the -data-dir argument earlier, but that was just me testing it out earlier to see if I got a different error message (in case consul WAS launching but having trouble reading that dir)...but no, still same result.

you left off the -u consul ?

Yeah, it works when I run it as root.

and what happens when you use that option?

That's what I'm currently doing, but it should be running as an unprivileged user.

daemon doesn't appear to be able to find /usr/local/bin/consul.

even when you run it yourself rather than via rc.subr ?

The port is sysutils/consul, which is version 1.12.4. I updated it internally to the latest version, 1.15.2.

The main port runs it as consul:consul by default.

Yeah, let me confirm that by adding -u from the shell.

It fails when I directly try run the above command from a root shell WITH `-u consul`, without that argument it runs fine (but it runs as root).

fails with the same error?

If /usr/local/bin/consul was needing some root permissions or whatnot, then I'd deal with that and if necessary I'd run it as root...but I'd expect the executable itself to complain about that, not the daemon process that's supervising it erroring out by saying "daemon: /usr/local/bin/consul: No such file or directory"

Yeah, same error. No such file or directory.

ok, try the same command but add ktrace -i in front, then run kdump and look near the end of the output for the error message

k one sec

then pastebin the preceding few hundred lines

or… dtruss?

I've never used kdump before. Can I clear whatever it's buffering/dumping it to and re-run that? I can't tell what's from this run vs previous runs (if it's still retaining that info)

ktrace writes to a file ktrace.out, kdump just reads that file

I see.

nothing is retained between runs

Yup, I see ktrace.out in cwd.

Let me termbin it.

but the start of the dump is a whole ton of boilerplate stuff just to start the program (if it's dynamic)

Yeah, looking for the right breakpoint to cut it. Starting at tail and scrolling up right now.

you could look for the fork, and start there.

(everything interesting to us should happen after the fork)

Heh, forgot it's a binary file and tried to read it directly.

you can kdump >blah.txt and then work with that

Yup that's what I'm doing now. Just want to make sure there's no sensitive info in it before I termbin it.

sure

Ok, I snipped what looked like a few blocks of environment read calls. I can probably put those back in there if they're relevant, one sec...let me termbin this, and the while it's being looked over and I'll do a re-review of the environment and re-insert those after I make sure there are no passwds/tokens/etc in there.

ok