Splitting out some of the weird issues | Dagger | Page 1

#

With @eager osprey this morning, I managed to get the issue reproducing locally, no longer need to keep pushing to CI to get it to happen

#

I also think I managed to do a couple of interesting things:

Detached this from the new plain progress
Detached this from python
https://github.com/jedevc/dagger/tree/fix-weird-stall/examples/sdk/go/repro

#

This is a commit right on top of main that gets the issue to happen in the go sdk - it simulates the exact scenario in python, and seems to be a telemetry draining issue

#

Weirdly, with my little python repro, I can manage to "fix" the issue by printing out lots of extra output to stdout/stderr - something like adding this to the top of session.go

    go func() {
        time.Sleep(2 * time.Second)
        for {
            fmt.Fprint(os.Stdout, ".")
            time.Sleep(100 * time.Millisecond)
        }
    }()

#

this "seems" to work for a fun reason - if you add this with the above go reproducer, you get this:

panic: signal: broken pipe

goroutine 1 [running]:
main.main()
    /home/jedevc/Documents/Projects/dagger/examples/sdk/go/repro/main.go:34 +0x185
exit status 2

#

so somehow - a pipe is closed, and the close completes

#

I'm really reaching the end of my level of understanding here sadly - I think there's something related to something being captured by OTEL, in a fairly unexpected way.

#

cc @verbal hamlet @split robin @eager osprey 🙂

#Splitting out some of the weird issues