runtime: gdb fails to wait for new child

[prattmic]

#!watchflakes
post <- pkg == "runtime" && test == "TestGdbBacktrace" && (`waiting for new child: No child processes`)

https://build.golang.org/log/c0a5543211386b919f5b5311ccadc8afa54cd79e

--- FAIL: TestGdbBacktrace (2323.65s)
    runtime-gdb_test.go:83: gdb version 10.1
    runtime-gdb_test.go:468: GDB command timed out after 38m43.017693938s: /usr/bin/gdb -nx -batch -iex add-auto-load-safe-path /workdir/go/src/runtime -ex set startup-with-shell off -ex break main.eee -ex run -ex backtrace -ex continue /workdir/tmp/TestGdbBacktrace2238907183/001/a.exe
    runtime-gdb_test.go:473: gdb output:
        Loading Go Runtime support.
        Breakpoint 1 at 0x80a2830: file /workdir/tmp/TestGdbBacktrace2238907183/001/main.go, line 17.
        [New LWP 532860]
        [New LWP 532862]
        [New LWP 532864]
        [New LWP 532865]

        Thread 1 "a.exe" hit Breakpoint 1, main.eee (~r0=<optimized out>) at /workdir/tmp/TestGdbBacktrace2238907183/001/main.go:17
        17	func eee() bool { return true }
        #0  main.eee (~r0=<optimized out>) at /workdir/tmp/TestGdbBacktrace2238907183/001/main.go:17
        #1  0x080a280f in main.ddd (~r0=<optimized out>) at /workdir/tmp/TestGdbBacktrace2238907183/001/main.go:14
        #2  0x080a27da in main.ccc (~r0=<optimized out>) at /workdir/tmp/TestGdbBacktrace2238907183/001/main.go:11
        #3  0x080a27aa in main.bbb (~r0=<optimized out>) at /workdir/tmp/TestGdbBacktrace2238907183/001/main.go:8
        #4  0x080a277a in main.aaa (~r0=<optimized out>) at /workdir/tmp/TestGdbBacktrace2238907183/001/main.go:5
        #5  0x080a286a in main.main () at /workdir/tmp/TestGdbBacktrace2238907183/001/main.go:22
        waiting for new child: No child processes.
    runtime-gdb_test.go:488: gdb exited with error: signal: killed

It seems to be this issue: https://github.com/bminor/binutils-gdb/blob/f38f7ef965553d5f8a78eba934bb9b192be39c61/gdb/linux-nat.c#L1857. After a clone, gdb fails to find the new child. It seems that either the kernel already killed it, or there is a race where PTRACE_EVENT_CLONE is sent before the child is waitable.

[gopherbot]

Change https://go.dev/cl/499975 mentions this issue: runtime: update skips for TestGdbBacktrace

[bcmills]

Is there an upstream issue for this? The only similar one I could find is https://sourceware.org/bugzilla/show_bug.cgi?id=7166, but that seems too old to be plausible.

(I did see https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=730444, but it includes an assertion failure that isn't present in the linked test log.)

[prattmic]

Not that I could find. We could file one I suppose, but we should probably try to get a repro first (this has only happened twice).