Compiling for maverick 10.10 x86_64 about rethinkdb HOT 14 CLOSED

Hi gysing,

Thanks for reporting the compilation error; we'll look into it.

You might have better luck with a higher version of gcc (4.6 should work) or removing -Werror in the Makefile, but we don't currently support 10.04 or 10.10 because the kernel is apparently missing some features we use.

from rethinkdb.

Specifically, the main feature missing is TCP_USER_TIMEOUT kernel option. We can work around it in the server, but we haven't done it yet, and it's a fairly significant amount of work. There are also some other minor annoyanced (gcc errors, libv8).

from rethinkdb.

It looks like TCP_USER_TIMEOUT was added in kernel version 2.6.37, which means ubuntu 11.04+

from rethinkdb.

I'm having the same issue in CentOS 6.3 (the latest as of right now). The kernel is 2.6.32-279.el6.x86_64. So I guess RPM systems would be out until a new kernel 2.6.37+ comes out as an update. You could roll your own kernel but RPM based systems probably aren't going to do that.

Is there some other RPM distro you were targeting?

from rethinkdb.

No, we currently don't have RPM support and don't support kernels before 2.6.37 -- sorry :( We're having lots of portability discussions right now, so we'll have a concrete idea and schedule of what we will and will not support soon.

from rethinkdb.

Ok, this is going into next. Assigning to @Tryneus. We have to build application-level heartbeats to make this happen.

from rethinkdb.

I wonder if these old files from the replication code could be of help... They don't do anything too fancy, just provide some convenient way of temporarily suspending the heartbeat watchdog (e.g. while large chunks of data are being transmitted over the connection and heartbeats wouldn't have a chance to get through within reasonable time).

https://github.com/rethinkdb/rethinkdb/blob/eb3e7215bc9b122a01fc2cbeae65e58384380eb1/src/replication/heartbeat_manager.hpp
https://github.com/rethinkdb/rethinkdb/blob/eb3e7215bc9b122a01fc2cbeae65e58384380eb1/src/replication/heartbeat_manager.cc

from rethinkdb.

@coffeemug -- can we support 10.04/10.10 now that we have application-level heartbeats?

from rethinkdb.

@mlucy -- we still need to verify that there aren't strange compiler bugs. As we just discussed, this bug is now in your hands.

from rethinkdb.

Reassigning to @Tryneus -- we should close this as soon as heartbeat code goes into next and we verify that rethink compiles and runs on 10.04. Marc -- you can ping Frank about the compilation since he's been working on that.

from rethinkdb.

Well, the new heartbeat is in, as of commit 929c836. I'll let Frank know.

from rethinkdb.

Can this issue be closed? We compile on Maverick successfully, and the more gcc-version specific warnings will be dealt with in 4.4. Does anyone object to closing this?

from rethinkdb.

I tried compiling on RHEL 6.3 from the next branch about a week ago and it failed with the same error. If that is expected behavior then close and I'll open a new issue. If the heartbeart was to enable older kernels, please leave open.

from rethinkdb.

@squarism -- talked to some people, rethink now compiles on maverick in a branch, awaiting code review before it can be merged into next. Issue #121 will address the RHEL/CentOS compilation issues. The current branch turns off warnings with a pragma -- if you do that, you should be able to build on RHEL (you might also need to pass LEGACY_LINUX=1 and NO_EPOLL=1 to Make). Closing this one.

from rethinkdb.