Cross compile telegraf depend on libc.so instead of libc.so.6 about telegraf HOT 7 CLOSED

Hi,

Your build file is building with CGO enabled, which is not what we support or do for any of our official builds:

CGO_ENABLED=1

By building with CGO enabled, you building a dynamic binary which will depend on external dependencies rather than a static binary.

from telegraf.

@powersj
I don't know much about golang, I guess in traditional sense, you've meant static linkage?. By "CGO_ENABLED=1" is dynamic linkage?.

And there is no workaround regarding this ?.

from telegraf.

I guess in traditional sense, you've meant static linkage?. By "CGO_ENABLED=1" is dynamic linkage?.

Yes, there are a few other consequences as well, but that is the general difference. You can see:

• https://pkg.go.dev/cmd/cgo
• https://stackoverflow.com/questions/61515186/when-using-cgo-enabled-is-must-and-what-happens
• https://medium.com/@pengcheng1222/exploring-cgo-enabled-in-go-23cf5cf2fe88

And there is no workaround regarding this ?.

Why do you need to build a dynamic binary? By generating a static binary you avoid this all together.

from telegraf.

The go app that are used in OpenWrt are mostly/all dynamic i guess, "CGO_ENABLED=1" is hardcoded in golang compilation supports on OpenWrt.

For compiling telegraf with static linkage, what do I need to do manually?, what is the command line?, does passing "CGO_ENABLED=0" is enough?, it doesn't seems to work for my case.

What is the complete command lines , other than simple "make build" ?.

from telegraf.

The go app that are used in OpenWrt are mostly/all dynamic i guess, "CGO_ENABLED=1" is hardcoded in golang compilation supports on OpenWrt.

I thought someone was already building openwrt builds before. Did something change such that it broke?

other than simple "make build" ?.

That is how we produce all our artifacts. That roughly translates to:

branch := $(shell git rev-parse --abbrev-ref HEAD)
commit := $(shell git rev-parse --short=8 HEAD)
INTERNAL_PKG=github.com/influxdata/telegraf/internal
LDFLAGS := $(LDFLAGS) -X $(INTERNAL_PKG).Commit=$(commit) -X $(INTERNAL_PKG).Branch=$(branch)
CGO_ENABLED=0 go build -ldflags "-w -s $(LDFLAGS)" ./cmd/telegraf

You can add additional GO flags if required, like GOOS, GOARCH, etc. Or add LDFLAGS

from telegraf.

I am not a OpenWrt developer, only a user, but a PR that's comes in here .

$(1)/usr/bin/telegraf config > $(1)/etc/telegraf.conf

The compilation is going fine for the targeted devices but It's only for getting telegraf.conf, it's problematic if the libc or target is different than build system, I have tested that PR, if i am building targeting for glibc x86_64 then of course binary running fine.

Thank you for those snippet, it will be coming handy if just like old time, there is already extracted or given telegraf.conf somewhere, other than needs to run telegraf binary first.

from telegraf.

$(1)/usr/bin/telegraf config > $(1)/etc/telegraf.conf

Keep in mind the user, you, can generate this later on. The configuration that is produced is not necessary to run telegraf as it is incomplete and does not include a working output. If a user needs this they can run the command and then go through it.

Our own builds will run go run ./cmd/telegraf config > etc/telegraf.conf to generate the full config. This avoids running the locally built binary which as you discovered may be built for a different architecture.

Hopefully all these various steps help. I am going to close this as I don't think this is actually an issue with Telegraf itself.

from telegraf.