Arch Linux User Repository

I wonder if anyone can give me any tips or insight before I try to report my problem to upstream Snapcast.

I've been running Archlinux (ARM) and compiling Snapcast on my 1GB memory humble little Raspberry Pi 2 1.1 (32 bit armv7h) for 2 years. Until version 0.26 had no problems. It always took aaaaaaages to compile this C++ thingy, but in the end it bravely managed to do it. Running MPD and Snapcast with Opus encoding was also very smooth, and the CPU and RAM has always been more than enough.

With the latest version 0.26 in December, I'm stuck. I can't finish compiling it, without running out of memory apparently. Tried already several times. It's the exact same system I've been running for 2 years. But somehow I cant get past 15% in the compilation process and makepkg aborts with:

boost::asio::wait_traits<std::chrono::_V2::steady_clock> > >::heap_entry> > >’ changed in GCC 7.1
 1198 |    _M_realloc_insert(end(), __x);
      |    ~~~~~~~~~~~~~~~~~^~~~~~~~~~~~
{standard input}: Assembler messages:
{standard input}:178970: Error: invalid operands (*UND* and .ARM.extab sections) for `-'
c++: fatal error: Killed signal terminated program cc1plus
compilation terminated.
make[2]: *** [server/CMakeFiles/snapserver.dir/build.make:118: server/CMakeFiles/snapserver.dir/control_session_http.cpp.o] Error 1
make[2]: Leaving directory '/home/r1/deleteme/snapcast/src/snapcast-0.26.0/build'
make[1]: *** [CMakeFiles/Makefile2:160: server/CMakeFiles/snapserver.dir/all] Error 2
make[1]: Leaving directory '/home/r1/deleteme/snapcast/src/snapcast-0.26.0/build'
make: *** [Makefile:136: all] Error 2
make: Leaving directory '/home/r1/deleteme/snapcast/src/snapcast-0.26.0/build'
==> ERROR: A failure occurred in build().
    Aborting...

Sometimes I get a kernel "oom kill" thingy. The system runs out of memory, and the kernel aborts the process and reboots.

alarmpi kernel:

cc1plus invoked oom-killer: gfp_mask=0x100cca(GFP_HIGHUSER_MOVABLE), 
CPU: 1 PID: 16110 Comm: cc1plus Tainted: G         C        5.10.83-1-r>
kernel: Hardware name: BCM2835
[<8010efe8>] (unwind_backtrace) from [<8010ade0>] (show_stack+0x10/0x14

and

oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allow>
Out of memory: Killed process 16110 (cc1plus) total-vm:876596kB, anon-r>
oom_reaper: reaped process 16110 (cc1plus), now anon-rss:0kB, file-rss:>

I have very little understanding of C++, compilation processes etc. But the first thing that occurred to me, is that, with version 0.26, maybe somehow the requirements of Snapcast have changed, are beefier now, and this hardware can't cope with it now.

Tried to shut down all the other non necessary services, and tried over and over to no avail.

Is this a plausible explanation ?

But before I report this to Snapcast upstream, I want to be 100% sure, that it's not the PKGBUILD fault.

I noticed something with this PKGBUILD here on AUR. The latest changes to the PKGBUILD removed this line:

- "gcc11-fix.patch"

So this patch is no longer being applied as was in version 0.25. I dont know what the patch was for but:

Could it somehow be related to this patch (or lack of it)  ?

Thanks in advance.

PS: Maybe related ? I saw this on Snapcast issues:

About the debian armhf package, https://github.com/badaix/snapcast/issues/960

...So the gcc and c++ libs are still needed in a more recent version. I guess because Snapcast 0.26
requires C++17 instead of C++11.

@cgirard: I've been trying to get exact solution upstreamed for quite a while now. See upstream issue: https://github.com/badaix/snapcast/issues/486

You're probably right that, in the meantime, I should change the text in the install file to refer to changing the location of the fifo rather than trying to work around the problems with the default location in /tmp

Hi, instead of advising of disabling protected FIFOs, it would be probably a better idea to just advise to use /run/snapserver/snapfifo as you even care to create /run/snapserver

@Hazzl: Thank you for upstreaming!

Since gcc11 hit the core repository today and no new release is available upstream (yet), I've added your patch. Thanks again for raising this issue.

@Hazzl: Could you please report this issue upstream?

I need the following patch to compile with g++-11.1.0

--- snapcast-0.25.0/common/message/message.hpp  2021-05-15 13:16:03.000000000 +0200
+++ snapcast-0.25.0b/common/message/message.hpp 2021-05-16 16:15:34.243593726 +0200
@@ -29,6 +29,7 @@
 #include <sys/time.h>
 #endif
 #include <vector>
+#include <memory>

 /*
 template<typename CharT, typename TraitsT = std::char_traits<CharT> >

@m040601: done

At https://github.com/badaix/snapcast there is the "docs" folder . Its content ends up in /usr/share/doc/snapcast

But there's also some extra usefull stuff besides that "docs" folder.

• A usefull file "changelog.md" file.
• A folder "control" containing example files using curl and jq

Could you please also include it ? Thanks.

Indeed, I just noticed that the original github alert is referring to a different commit than the current 0.22.0 tag. It seems he added one more commit to fix a missing include, and then re-tagged. Glad to have that cleared up. I thought I was going crazy. :-)

Probably happened because snapcast dev pulled v0.22.0 assets about an hour after they uploaded it to github, then uploaded new assets under the same release tag a bit later.