building on illumos gets stuck occasionally #35261
Labels
Comments
|
I believe the build in progress is: |
|
It seems like this might have been caused by the integration of 6becb03. If I try to build that revision, we seem to hit a reproducible hang at the @ianlancetaylor Sorry to bug you, but do you have any thoughts? |
|
When this happens again, can you send a |
|
I waited until the thing was stuck for at least a minute with no child processes and then I dropped |
|
Alright I think I have figured it out: diff --git a/src/runtime/netpoll_solaris.go b/src/runtime/netpoll_solaris.go
index 26bbe38d86..c05c2a2a7c 100644
--- a/src/runtime/netpoll_solaris.go
+++ b/src/runtime/netpoll_solaris.go
@@ -230,7 +230,13 @@ func netpoll(delay int64) gList {
retry:
var n uint32 = 1
if port_getn(portfd, &events[0], uint32(len(events)), &n, wait) < 0 {
- if e := errno(); e != _EINTR && e != _ETIME {
+ e := errno()
+ // As per port_getn(3C), an ETIME failure does not preclude the
+ // delivery of some number of events.
+ if e == _ETIME && n > 0 {
+ goto process
+ }
+ if e != _EINTR && e != _ETIME {
print("runtime: port_getn on fd ", portfd, " failed (errno=", e, ")\n")
throw("runtime: netpoll failed")
}
@@ -242,6 +248,7 @@ retry:
goto retry
}
+process:
var toRun gList
for i := 0; i < int(n); i++ {
ev := &events[i]I'll submit a CL. Sorry for the bother. |
|
Thanks for figuring this out! (Probably the real CL should not use a |
|
Change https://golang.org/cl/204801 mentions this issue: |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
As I've been trying to look into #35085, I've noticed the builder become stuck a couple of times. Most recently I've dug in to see where.
The last build action:
This has been running for some time:
The process tree:
There's an OS thread possibly stuck in the allocator?
The text was updated successfully, but these errors were encountered: