sjorge: FWIW, I was able to clone the repository and just "gmake" on a relatively stock OmniOS machine with development tools installed

(we use it for some of the illumos.org stuff)

I continously get "note: Not enough space (os error 12)" while building fish on OmniOS in a zone. There is no quotes set up for this dataset and the pool have enough free space. What can be the reson for this error?

ENOMEM 12

so, you are running out of memory

Aha!

Fck, I got only 16 Gigs.

os errors are listed in /usr/include/sys/errno.h

MAybe I should switch of some zones

right, I should have checked, but the error message was misleading.

not enough space in terms of memory is about virtual memory, thats ram + swap space (in case the allocations do reserve page space from swap).

maybe your problem is related to ulimits or resource limits on your zones. not sure what "fish" is and how big it is, but 16G seem plenty.

and virtual memory space may be eaten away with usage of /tmp -- some builds can drop large amount of data in /tmp

unless you are short of disk space, I'd give some 16G to swap to avoid surprises.

thanks for the hints, I will check and fix this.

Woodstock: fish is a non-posix shell, written in rust

for some reason the recent 2 fish version compiles fine, but if I attempt to run it, it drops me out of the zone.

No error, no core dump.

getting SIGHUP?

szilard@build:~/projekt/fish-4.5.0/b$ ./fish

[Connection to zone 'build' pts/9 closed]

root@omnios:~#

I am starting it from bash, but setting it as login shell causes the same, so it is not bash related.

4.3.x works, but 4.4 and onwards is broken somewhere.

truss -fo truss.txt ./fish and then see whats in truss.txt ;)

This is 4.5.0, which is strange, because this version is advertised as illumos compatible.

it is long, let me paste it somewhere.

tsoome: paste.omnios.org/?fcf819cd274171f5#…F18Qescp5vc7qGZrA2AFxSxrCSpkMg8r2nf

17365/1: Received signal #1, SIGHUP, in pollsys() [caught]

it doesn1t tells too much, but at least I can paste this into a ticket upstream.

also note the write and ioctl calls with fd 0, 1, 2 are failing (it does not have access to stdin, stdout, stderr).

note, you get those signals after this write: write(1, "1B [ ? u1B [ > 0 q1B ] 1".., 83) = 83 -- it smells like terminal CSI sequence with printing 1B instead of ESC (0x1b) and then performing some error out procedure.

I pasted your comment aswell into the ticket, hope it is ok.

its only a guess based on the trace, but thats where things seem to happen and therefore something must be fishy there;)

:P

I have this:

file path=opt/ooce/include/event.h owner=root group=bin mode=0644 facet.devel=true

but I should move this file into include/event2/ fodler. Can I do this with local.mog?

szilard: I'd look closely at the ioctls immediately before the error - something job control related, perhaps..

I failed to compile the latest zabbix :(

but I just noticed there is pre-built prometheus available on github

what problems are you seeing with the zabbix build?

It needs libevent. I have compiled it, but then zabbix's configure fails with missing symbols in my libevent.

For zabbix I use the same mechanism as for some other applications and build a temporary static libevent (2.1.12 if it matters) and point the build at that.

This is my attempt on libevent: omniosorg/omnios-extra #1712

I took inspiration from the pkgsrc zabbix recipe (however thats a much older variant)

err, I mean they also have the recent libevent, but their zabbix version is old

Other thing: I came up with this SMF xml for prometheus's node-exporter: paste.omnios.org/?e46915fc941d0565#…qqCFTe5Jz6DqvQz8Psn48qDncVtA1tqqyyy

I have imported and enabled it, but while it starts node-exporter, it shows the svc as offline.

actually it is offline*, next state supposed to be online, but it never switches to online.

Can somebody give me a hint, how to fix it?

check service dependencies svcs -d and -D

you probably are missing its config file or like...

nothing obvious: paste.omnios.org/?59af1a2e45614d5e#…8gUhVAcvnyzEN9PN7e6ASLKrPY9WkVWRF5L

looks like its starting. anything in logs?

according to the logs it runs fine.

svcs -xv gives this hint illumos.org/msg/SMF-8000-C4

Couple of differences from my smf manifest:

In the start method section, I have timeout_seconds set to 60 not 0

In the startd framework, I have a duration of child, not contract

ptribble: with the mentioned changes and deleting/reimporting and enabling the svc I got "uninitialized" state

but reducing the timeout again to 0 while keeping the "child" i got it online.

As for the build, I would expect it to just work, and just tested 7.0.23 and 7.4.7 on Tribblix with a temporary libevent and they both built cleanly

I probably did something wrong with my libevent.

I see this:

configure:14487: checking for libevent support

configure:14676: gcc -o conftest -m64 -g -O2 -DZBX_OLD_SOLARIS -I/opt/ooce/include -rdynamic -L/opt/ooce/lib/amd64/ conftest.c -lpthread -lkvm -lm -lkstat -lsocket -lresolv -lev >&5

Undefined first referenced symbol in file

evthread_use_pthreads /var/tmp//ccvxayma.o

and

szilard@build:~/projekt/zabbix-7.0.23/b$ objdump -x /opt/ooce/lib/amd64/libevent_pthreads.so |grep evthread_use_pthreads

0000000000001b20 g F .text 00000000000000be evthread_use_pthreads

that link command line is missing -levent_pthreads

I know. I have tried to prefix the configure invocation with LDFLAGS="-levent_pthreads" but the lib is not in the searched directory, so I have attempted to add it as -L/opt/ooce/lib/amd64 but failed every time, no idea why really.

but it is getting late here, need to have some rest, I'll look into it tomorrow.

Something's awry, because that configure check you show is looking for libev not libevent