[Bug 211871] poll() is probably broken in 11.0-RC1
bugzilla-noreply at freebsd.org
bugzilla-noreply at freebsd.org
Mon Aug 15 15:51:19 UTC 2016
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=211871
Bug ID: 211871
Summary: poll() is probably broken in 11.0-RC1
Product: Base System
Version: 11.0-RC1
Hardware: Any
OS: Any
Status: New
Severity: Affects Some People
Priority: ---
Component: kern
Assignee: freebsd-bugs at FreeBSD.org
Reporter: afiskon at devzen.ru
Hello.
I'm a contributor to PostgreSQL project. After upgrading FreeBSD from
11.0-BETA4 to 11.0-RC1 I noticed something very weird. Here are steps to
reproduce:
```
sudo pkg install git-lite bison perl5 gmake
git clone git clone http://git.postgresql.org/git/postgresql.git
cd postgresql
# for the record - I'm on 9389fbd0
./configure --enable-tap-tests --enable-cassert --enable-debug
gmake -j4 -s
gmake check
```
On 11.0-RC1 `gmake check` hangs like this:
```
returning ... ok
largeobject ... ok
with ... ok
xml ... ok
test event_trigger ... ok
test stats ...
```
This issue reproduces in 100% of cases. From what I can tell (judging by DTrace
and stacktraces in gdb) it looks like poll() was broken. Sometimes it doesn't
see changes.
As a side note you shouldn't probably believe me on bare word here since I'm
not familiar with this particular part of PostgreSQL and I have no experience
in debugging a kernel. The real issue could be somewhere else.
Naturally it could be a bug in PostgreSQL's code. But it works flawlessly not
only on Linux distributions but also on 10.3 (I've just re-checked). There is
currently a feature freeze in PostgreSQL project. So there were no major
changes in a code recently (like three months). Thus I believe it's most likely
a bug in a kernel.
--
You are receiving this mail because:
You are the assignee for the bug.
More information about the freebsd-bugs
mailing list