Scylladb: fix build under sandboxing #70576

workflow · 2019-10-07T05:32:45Z

Motivation for this change

Build was broken under sandboxing.
This fixes two build nondeterminisms

CC: @clojurians-org @erictapen

Resolves #70049

Also ticks one of the boxes in #68361

Things done

Co-authored-by: Justin Humm <justin.humm@posteo.de>

workflow · 2019-10-07T05:36:14Z

@GrahamcOfBorg build scylladb

erictapen · 2019-10-07T14:05:02Z

Did you consider upstreaming the patch? I guess this would be a good change for them as my os-release man page says, that ID can default to just linux.

Besides that, this LGTM but as I wrote part of it I guess it makes sense that somebody else approves these changes.

jonringer · 2019-10-07T15:07:55Z

@GrahamcOfBorg build scylladb

jonringer · 2019-10-07T15:08:53Z

I had trouble building this:

Build Failure

builder for '/nix/store/k2npwcd30nrcbgyy8xl67mqrn1pjwh2i-scylladb-3.0.5.drv' failed with exit code 1; last 10 log lines:
  g++ -MD -MT build/release/database.o -MF build/release/database.o.d -std=gnu++1z -g  -Wall -Werror -Wno-error=deprecated-declarations -fvisibility=hidden  -pthread -I/build/scylla-403f66e/seastar -U_FORTIFY_SOURCE  -gz -march=westmere -fconcepts -I/build/scylla-403f66e/seastar/fmt -DBOOST_TEST_DYN_LINK -Wno-ignored-attributes -Wno-overloaded-virtual -Wno-maybe-uninitialized -Wno-error=cpp -Wno-stringop-overflow -DSEASTAR_HAVE_GCC6_CONCEPTS -DSEASTAR_HAS_MEMBARRIER -DSEASTAR_HAVE_LZ4_COMPRESS_DEFAULT -DSEASTAR_HAVE_ASAN_FIBER_SUPPORT -O2 -O3 --param inline-unit-growth=300 -I/build/scylla-403f66e/seastar/build/release/gen -I/build/scylla-403f66e/seastar/build/release/c-ares -I/nix/store/zf9b2449vdif958ka250i5h0qh5kkzan-jsoncpp-1.9.1/include -DBOOST_TEST_DYN_LINK -I/nix/store/76a70m24s0rkz3id21pipqz9g669qwq4-systemd-243-dev/include -Wno-maybe-uninitialized -Wno-tautological-compare -Wno-missing-braces -Wno-misleading-indentation -Wno-overflow -Wno-noexcept-type -Wno-nonnull-compare -Wno-error=deprecated-declarations -DHAVE_LIBSYSTEMD=1 -O3 --param inline-unit-growth=300 -DXXH_PRIVATE_API -I. -I build/release/gen -I seastar -I seastar/build/release/gen  -c -o build/release/database.o database.cc
  g++: fatal error: Killed signal terminated program cc1plus
  compilation terminated.
  [60/577] CXX build/release/cql3/statements/create_table_statement.o
  FAILED: build/release/cql3/statements/create_table_statement.o
  g++ -MD -MT build/release/cql3/statements/create_table_statement.o -MF build/release/cql3/statements/create_table_statement.o.d -std=gnu++1z -g  -Wall -Werror -Wno-error=deprecated-declarations -fvisibility=hidden  -pthread -I/build/scylla-403f66e/seastar -U_FORTIFY_SOURCE  -gz -march=westmere -fconcepts -I/build/scylla-403f66e/seastar/fmt -DBOOST_TEST_DYN_LINK -Wno-ignored-attributes -Wno-overloaded-virtual -Wno-maybe-uninitialized -Wno-error=cpp -Wno-stringop-overflow -DSEASTAR_HAVE_GCC6_CONCEPTS -DSEASTAR_HAS_MEMBARRIER -DSEASTAR_HAVE_LZ4_COMPRESS_DEFAULT -DSEASTAR_HAVE_ASAN_FIBER_SUPPORT -O2 -O3 --param inline-unit-growth=300 -I/build/scylla-403f66e/seastar/build/release/gen -I/build/scylla-403f66e/seastar/build/release/c-ares -I/nix/store/zf9b2449vdif958ka250i5h0qh5kkzan-jsoncpp-1.9.1/include -DBOOST_TEST_DYN_LINK -I/nix/store/76a70m24s0rkz3id21pipqz9g669qwq4-systemd-243-dev/include -Wno-maybe-uninitialized -Wno-tautological-compare -Wno-missing-braces -Wno-misleading-indentation -Wno-overflow -Wno-noexcept-type -Wno-nonnull-compare -Wno-error=deprecated-declarations -DHAVE_LIBSYSTEMD=1 -O3 --param inline-unit-growth=300 -DXXH_PRIVATE_API -I. -I build/release/gen -I seastar -I seastar/build/release/gen  -c -o build/release/cql3/statements/create_table_statement.o cql3/statements/create_table_statement.cc
  g++: fatal error: Killed signal terminated program cc1plus
  compilation terminated.
  [77/577] CXX build/release/main.o_test_env.oions.o
  ninja: build stopped: subcommand failed.
cannot build derivation '/nix/store/ga97vinhphb6gl5dqlbsxh6f0xmsbpl4-env.drv': 1 dependencies couldn't be built
[0 built (1 failed)]
error: build of '/nix/store/ga97vinhphb6gl5dqlbsxh6f0xmsbpl4-env.drv' failed
https://github.com/NixOS/nixpkgs/pull/70576
1 package failed to build:
scylladb

No packages were successfully build, skip nix-shell

workflow · 2019-10-09T01:45:59Z

I had trouble building this:

Build Failure

@jonringer Thanks for triggering the build! This is a really heavy one, and it looks like grahamcofborg ran out of RAM. Scrolling up a bit in the logs:

virtual memory exhausted: Cannot allocate memory
[77/577] CXX build/release/tests/cql_test_env.o
FAILED: build/release/tests/cql_test_env.o
g++ -MD -MT build/release/tests/cql_test_env.o -MF build/release/tests/cql_test_env.o.d -std=gnu++1z -g  -Wall -Werror -Wno-error=deprecated-declarations -fvisibility=hidden  -pthread -I/build/scylla-403f66e/seastar -U_FORTIFY_SOURCE  -gz -march=westmere -fconcepts -I/build/scylla-403f66e/seastar/fmt -DBOOST_TEST_DYN_LINK -Wno-ignored-attributes -Wno-overloaded-virtual -Wno-maybe-uninitialized -Wno-error=cpp -Wno-stringop-overflow -DSEASTAR_HAVE_GCC6_CONCEPTS -DSEASTAR_HAS_MEMBARRIER -DSEASTAR_HAVE_LZ4_COMPRESS_DEFAULT -DSEASTAR_HAVE_ASAN_FIBER_SUPPORT -O2 -O3 --param inline-unit-growth=300 -I/build/scylla-403f66e/seastar/build/release/gen -I/build/scylla-403f66e/seastar/build/release/c-ares -I/nix/store/zf9b2449vdif958ka250i5h0qh5kkzan-jsoncpp-1.9.1/include -DBOOST_TEST_DYN_LINK -I/nix/store/76a70m24s0rkz3id21pipqz9g669qwq4-systemd-243-dev/include -Wno-maybe-uninitialized -Wno-tautological-compare -Wno-missing-braces -Wno-misleading-indentation -Wno-overflow -Wno-noexcept-type -Wno-nonnull-compare -Wno-error=deprecated-declarations -DHAVE_LIBSYSTEMD=1 -O3 --param inline-unit-growth=300 -DXXH_PRIVATE_API -I. -I build/release/gen -I seastar -I seastar/build/release/gen  -c -o build/release/tests/cql_test_env.o tests/cql_test_env.cc
virtual memory exhausted: Cannot allocate memory
[78/577] CXX build/release/database.o

How do we resolve that?

I could patch it to run on $NIX_BUILD_CORES / 2, which I've found to be much better behaved in many instances (likely due to hyperthreading), but not sure that should be in the scope of this PR.

workflow · 2019-10-09T01:50:28Z

@infinisil Can we trouble you to take a look?

jonringer · 2019-10-09T04:01:21Z

I did notice that the review almost maxed out my 32Gb :(

workflow · 2019-10-09T23:49:08Z

From the manual

Note: Compiling Scylla requires, conservatively, 2 GB of memory per native thread, and up to 3 GB per native thread while linking. GCC >= 8.1.1. is required.

...so I think in this case we could reduce the -j somehow accordingly to the available memory on the system. However I would advocate doing that in a separate PR, as this PR is already an incremental improval on the build.

workflow · 2019-10-09T23:52:34Z

Did you consider upstreaming the patch? I guess this would be a good change for them as my os-release man page says, that ID can default to just linux.

Besides that, this LGTM but as I wrote part of it I guess it makes sense that somebody else approves these changes.

@erictapen I did not yet, good idea. Please, go for it if you have time :)

vcunat · 2019-10-10T10:55:18Z

I noticed yesterday and used de51368. It would be possible to, say, use half of --cores, but such approaches seem relatively hacky. 2 GB seem still borderline-acceptable to me, as we have other builds eating a lot – when I started with Hydra's small nodes (t4a and t4b), I found that 8 GB was not enough for --max-jobs 4 --cores 4 in our case and 12-16 seemed OK (with /tmp/ outside RAM, at least).

jonringer · 2019-10-11T01:36:08Z

probably the safest way to go is enableParallelBuilding = false;, making any assumptions about $NIX_BUILD_CORES should be left up to the user to configure.

vcunat · 2019-10-11T17:26:35Z

Well, that's like setting cores to 1. Anyway, the main problem I see ATM is that the output is around 18 GiB, and we can't have that on the binary cache, meaning I'll disable it on Hydra as there's little point when it can't "succeed".

Perhaps almost all of the content is useless? (e.g. I see >10G in build/release/tests/) We have a hard-limit of 2G. This part is probably an overkill and messy anyway:

{
  installPhase = ''
    mkdir $out
    cp -r * $out/
  '';
}

jonringer · 2019-10-11T17:33:46Z

Agreed, that's too big.

(cherry picked from commits 3596c4c, c0d5833, 8ae02b5)

(cherry picked from commit 8ae02b5)

workflow and others added 3 commits October 7, 2019 13:21

scylladb: fix build access to /etc/os-release

b932511

Co-authored-by: Justin Humm <justin.humm@posteo.de>

scylladb: patch shebangs in further input file

7a4f480

scylladb: fix spacing in default.nix

bbaaefd

ofborg bot added 11.by: package-maintainer 10.rebuild-darwin: 0 10.rebuild-linux: 1-10 labels Oct 7, 2019

vcunat added a commit that referenced this pull request Oct 10, 2019

Merge #70576: scylladb: fix build under sandboxing

3596c4c

vcunat merged commit bbaaefd into NixOS:master Oct 10, 2019

vcunat added a commit that referenced this pull request Oct 12, 2019

scylladb: don't build on Hydra; /cc PR #70576

8ae02b5

vcunat added a commit that referenced this pull request Oct 12, 2019

Merge #70576: scylladb: fix build under sandboxing

6422227

(cherry picked from commits 3596c4c, c0d5833, 8ae02b5)

dtzWill pushed a commit to dtzWill/nixpkgs that referenced this pull request Jan 28, 2020

scylladb: don't build on Hydra; /cc PR NixOS#70576

772eca0

(cherry picked from commit 8ae02b5)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Scylladb: fix build under sandboxing #70576

Scylladb: fix build under sandboxing #70576

workflow commented Oct 7, 2019

workflow commented Oct 7, 2019

erictapen commented Oct 7, 2019

jonringer commented Oct 7, 2019

jonringer commented Oct 7, 2019

workflow commented Oct 9, 2019 •

edited

workflow commented Oct 9, 2019

jonringer commented Oct 9, 2019

workflow commented Oct 9, 2019 •

edited

workflow commented Oct 9, 2019 •

edited

vcunat commented Oct 10, 2019

jonringer commented Oct 11, 2019

vcunat commented Oct 11, 2019

jonringer commented Oct 11, 2019

Scylladb: fix build under sandboxing #70576

Scylladb: fix build under sandboxing #70576

Conversation

workflow commented Oct 7, 2019

Motivation for this change

Things done

workflow commented Oct 7, 2019

erictapen commented Oct 7, 2019

jonringer commented Oct 7, 2019

jonringer commented Oct 7, 2019

workflow commented Oct 9, 2019 • edited

workflow commented Oct 9, 2019

jonringer commented Oct 9, 2019

workflow commented Oct 9, 2019 • edited

workflow commented Oct 9, 2019 • edited

vcunat commented Oct 10, 2019

jonringer commented Oct 11, 2019

vcunat commented Oct 11, 2019

jonringer commented Oct 11, 2019

workflow commented Oct 9, 2019 •

edited

workflow commented Oct 9, 2019 •

edited

workflow commented Oct 9, 2019 •

edited