Prolog of hermetic build with GCC 11 and Clang 13. #7712

amosbird · 2022-01-11T10:58:29Z

Proposed changes

Prepare to generate hermetic build using GCC 11 and Clang 13. The ideal toolchain would be ldb toolchain generated by ldb_toolchain_gen.sh

To kick off a clang build, set DORIS_TOOLCHAIN=clang before running any build scripts.

Types of changes

What types of changes does your code introduce to Doris?
Put an x in the boxes that apply

Bugfix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
Documentation Update (if none of the other choices apply)
Code refactor (Modify the code structure, format the code, etc...)
Optimization. Including functional usability improvements and performance improvements.
Dependency. Such as changes related to third-party components.
Other.

Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your code.

I have created an issue on (Fix #ISSUE) and described the bug/feature there in detail
Compiling and unit tests pass locally with my changes
I have added tests that prove my fix is effective or that my feature works
If these changes need document changes, I have updated the document
Any dependent changes have been merged

Further comments

The experience of compiling Doris can be improved in many ways (I failed to compile it several times over the last year). The non-intrusive way of doing it will be providing a dedicated toolchain to build both thirdparty deps and the backend. This PR contains such proposal along with some improvements for building the thirdparty on unclean environment (some environment provides system shared libs which should not be found and used).

Benefits:

Generate Byte-identical binaries. Good for quality control, testing, debugging, redistributing and forking.
No docker dependency (your dev environment will appreciate that).
Extremely friendly to new developers and potentially attracting more contributors
Much easier to correctly maintain all build deps.

With the ldb toolchain, a minimal centos-7 dev machine (without web-ui things) will be the following:

    yum install -y \
        byacc \
        patch \
        automake \
        libtool \
        make \
        which \
        file \
        ncurses-devel \
        gettext-devel \
        bison \
        java-11-openjdk-devel \
        maven \
        unzip \
        bzip2 \
        zip

Why centos 7? Because it's the still maintained linux distribution with oldest GLIBC (2.17). https://gist.github.com/wagenet/35adca1a032cec2999d47b6c40aa45b1

It's also possible to enrich the LDB toolchain so that no external dependency is required.

It's worth mentioning if LDB toolchain is used, env ASAN_SYMBOLIZER_PATH should be set to the path of llvm-symbolizer explicitly.

With alternative toolchains, we can catch undefined behaviors like https://github.com/apache/incubator-doris/blob/a034c20d169d65b500bd984240eacd5c89c68571/be/src/util/simd/vstring_function.h#L93-L112

The undefined behavior of __builtin_ctz is described here https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html cc @HappenLee

New toolchain also indicates that sm3_test has potential mem leak which comes from openssl (maybe false positive).

Related PRs #7631 #7569

yiguolei · 2022-01-12T04:37:36Z

Great feature

yiguolei · 2022-01-12T06:22:43Z

LGTM

morningman · 2022-01-12T16:22:13Z

Link to #7590

morningman · 2022-01-11T14:02:05Z

be/src/glibc-compatibility/musl/eventfd.c

+{
+	int r = __syscall(SYS_eventfd2, count, flags);
+#ifdef SYS_eventfd
+	if (r==-ENOSYS && !flags) r = __syscall(SYS_eventfd, count);


tab to space

I'm not sure if we should change code style from external projects.

Oh, I see, the dir has been added to the .clang-format-ignore.
LGTM

So before merging this PR, we need to update the current docker dev image, so that developer can still compile from source, am I right?

we need to update the current docker dev image

Lemme check if it introduces any kinds of regressions. In general, this PR is for gcc11 and clang13 but should also play well with slightly older toolchains.

It seems apache/incubator-doris:build-env-latest has prebuilt thirparty blobs. This PR introduces some tweaks along with a new lib libidn. Perhaps it's a good time to rebuilt the dev image with new toolchains.

thirdparty/patches/googletest-release-1.11.0.patch | 71 ++++++++++ thirdparty/patches/libevent.patch | 11 ++ thirdparty/patches/rocksdb-5.14.2.patch | 11 ++

morningman · 2022-01-12T16:51:40Z

LGTM except for some small problem

yangzhg · 2022-01-13T11:37:20Z

thirdparty/build-thirdparty.sh

    --without-librtmp --with-ssl=${TP_INSTALL_DIR} --without-libidn2 --disable-ldap --enable-ipv6 \
-    --without-libssh2
-    make -j $PARALLEL && make install
+    --without-libssh2 --without-brotli


It's better to add --without-libpsl to avoid undefined reference to psl_free in some env

I think we can refine the thirdparty builds gradually in subsequent PRs. After we encourage hermetic build, there will be a broad range of environment for testing. Corner cases are definitely expected.

yangzhg · 2022-01-14T04:04:47Z

be/CMakeLists.txt

+    set(ASAN_LIBS -static-libasan)
+    set(LSAN_LIBS -static-liblsan)
+    set(UBSAN_LIBS -static-libubsan tcmalloc)
+    set(TSAN_LIBS -static-libtsan)


I am not sure which lib of asan, tsan,lasn,usan will be used, if the license of that lib is LGPL or other, GPL like license we may only allow to link the dynamic lib

we may only allow to link the dynamic lib

That's not the case. We already provide everything that allow the user to relink the application. Even our toolchain is open sourced :)

To be more specific, there is nothing in LGPL that mandates dynamic linking/loading. The requirement is that the end user should be able to relink the application to a substitute set of libraries if he so chooses.

amosbird · 2022-01-18T06:35:58Z

Now the PR supports CentOS 6 compilation.

NOTE: higher version of bison and autoconf is required.

NOTE2: plugin_loader_test ships a prebuilt shared object which contains GLIBC 2.14 memcpy. It cannot be loaded on CentOS 6 (GLIBC 2.12). This PR updates the blob (compiled with ldb-toolchain-v0.3 clang-13, linked with ld.bfd).

morningman

LGTM

github-actions · 2022-01-21T04:10:48Z

PR approved by at least one committer and no changes requested.

github-actions · 2022-01-21T04:10:51Z

PR approved by anyone and no changes requested.

…pache#7712)" This reverts commit 800a363.

morningman assigned morningman, yangzhg and HappenLee Jan 12, 2022

morningman added kind/chore Issues or PRs related to compile and install kind/improvement labels Jan 12, 2022

morningman reviewed Jan 12, 2022

View reviewed changes

amosbird force-pushed the build branch from f06a395 to b845ca8 Compare January 13, 2022 09:22

yangzhg reviewed Jan 13, 2022

View reviewed changes

yangzhg reviewed Jan 14, 2022

View reviewed changes

amosbird force-pushed the build branch from 31ff1f5 to 0b2714d Compare January 18, 2022 06:18

amosbird added 3 commits January 20, 2022 22:04

Prolog of hermetic build with GCC 11 and Clang 13.

6d14c05

Fix centos 6 build

496eec3

Fix warnings

e77b438

amosbird force-pushed the build branch from 0b2714d to e77b438 Compare January 20, 2022 16:17

More cleaning and revise

3d0bc5a

morningman approved these changes Jan 21, 2022

View reviewed changes

github-actions bot added the approved Indicates a PR has been approved by one committer. label Jan 21, 2022

github-actions bot added the reviewed label Jan 21, 2022

morningman merged commit 800a363 into apache:master Jan 21, 2022

zuochunwei pushed a commit to zuochunwei/incubator-doris that referenced this pull request Jan 21, 2022

Revert "[chore] Prolog of hermetic build with GCC 11 and Clang 13. (a…

9a473c0

…pache#7712)" This reverts commit 800a363.

morningman mentioned this pull request Feb 6, 2022

[Fix] Support compiling by gcc 11. #7631

Closed

amosbird mentioned this pull request Feb 7, 2022

[fix] fix ltrim result may incorrect in some case #7963

Merged

Prolog of hermetic build with GCC 11 and Clang 13. #7712

Prolog of hermetic build with GCC 11 and Clang 13. #7712

Uh oh!

Conversation

amosbird commented Jan 11, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Proposed changes

Types of changes

Checklist

Further comments

Uh oh!

yiguolei commented Jan 12, 2022

Uh oh!

yiguolei commented Jan 12, 2022

Uh oh!

morningman commented Jan 12, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

amosbird Jan 13, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

morningman commented Jan 12, 2022

Uh oh!

yangzhg Jan 13, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

yangzhg Jan 14, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

amosbird commented Jan 18, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

morningman left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 21, 2022

Uh oh!

github-actions bot commented Jan 21, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

amosbird commented Jan 11, 2022 •

edited

Loading

amosbird Jan 13, 2022 •

edited

Loading

yangzhg Jan 13, 2022 •

edited

Loading

yangzhg Jan 14, 2022 •

edited

Loading

amosbird commented Jan 18, 2022 •

edited

Loading