Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

MDBF-793 - Retire MSAN clang-15 builder, upgrade to clang-19 #562

Open
wants to merge 130 commits into
base: dev
Choose a base branch
from

Conversation

RazvanLiviuVarzaru
Copy link
Collaborator

@RazvanLiviuVarzaru RazvanLiviuVarzaru commented Sep 12, 2024

Replace MSAN CLANG-15 builder with CLANG 19 on Debian 12.

@RazvanLiviuVarzaru RazvanLiviuVarzaru marked this pull request as ready for review September 12, 2024 12:46
grooverdan and others added 12 commits September 19, 2024 15:21
Add worker for openeuler 24.03.

Rename container generation to mathc the same name.
Clients in our binary taballs are linked dynamically with libncurses, which causes a trouble on systems where it is no longer available
As per MDBF-794 only centos7 and debian 10 bintar builders need to link statically with libncurses.

- for debian 10: libncurses.a is already present in /scripts/local/lib/ and -DCURSES_NEED_NCURSES=1 flag is enough to link it statically.
- for centos7: I've rebuilt the image and pushed manually a dev_ tag. libncurses.a was present on the system but not in the library path: /scripts/local/lib/

Tests available at:
- debian10: https://buildbot.dev.mariadb.org/#/builders/187/builds/23
- centos7: https://buildbot.dev.mariadb.org/#/builders/172/builds/292

Reference CMAKE output line:
-- Found Curses: /scripts/local/lib/libncurses.a
Adjust images to 1505 notation as well.
@RazvanLiviuVarzaru
Copy link
Collaborator Author

@grooverdan
I think this closes #482.

Fedora 41 is including MariaDB 10.11 so
use that as the minimum version.
fauust and others added 12 commits September 30, 2024 13:11
Centos based have asio-devel in EPEL.

SLES is omitted as it requires a subpackage module
repository to be enabled. Doing so would increase
the package dependencies when using the built packages.
https://packagehub.suse.com/packages/asio/
(SLE-Module-PackageHub-Subpackages-Module )

OpenSUSE also has asio-devel but adds a complicated
path in with the openssl libraries and boost that
we'd be influcting on users if they used a non-bundled
version.

On RHEL - asio-devel in ubi9 EPEL but not 8.
Parts 1 and 2: Switching minor upgrade into "all" mode and extending
it with additional checks (ldd, requirements, capabilities)
During the "Compile step" - create_initial_db.cmake runs a bootstrap as a sanity check to avoid spending time on test runs.

It can happen that some changes on the server may cause a loop during the bootstrap process. Buildbot will terminate the parent process (dojob) when it reaches a timeout, but the MariaDB service will remain up causing new builds to fail.
https://buildbot.mariadb.org/#/builders/234/builds/32911

Modified the stop_processes step so handle this.
Use ( deb | archive ).mariadb.org to download the Packages file. This file is mandatory for both the 'All' and 'Columnstore' test types, exit with error if it is not found.

Apply uniq to remove duplicate values in the package list and sort the list for readability.
… old bb

- .build-id files may also be present on old systems where --noartifact
  does not work
- package dependencies need to be sorted for comparison
- dependency collection hangs frequently in buildbot, enabled logging
  temporarily to get more output for the issue
… old bb

- .build-id files may also be present on old systems where --noartifact
  does not work
- package dependencies need to be sorted for comparison
- dependency collection hangs frequently in buildbot, enabled logging
  temporarily to get more output for the issue
Improving build time by excluding the too many files of mariadb-test package from the ldd check when dependencies are collected.
And the --force is for debian:11 compat where the files exist in the tarball.
Didn't instrument sufficiently.

MSAN errors resultin the form:

==9000==WARNING: MemorySanitizer: use-of-uninitialized-value
[Detaching after fork from child process 9003]
    #0 0x55c87cbda03f in my_base64_encode /source/mysys/base64.c:113:13
    #1 0x55c87af96efb in calculate_server_uid(char*) /source/sql/mysqld.cc:9889:3
    #2 0x55c87af96efb in init_common_variables() /source/sql/mysqld.cc:3882:7
    #3 0x55c87af9111e in mysqld_main(int, char**) /source/sql/mysqld.cc:5546:7
    #4 0x7f6d94ab5249  (/lib/x86_64-linux-gnu/libc.so.6+0x27249) (BuildId: c047672cae7964324658491e7dee26748ae5d2f8)
    MariaDB#5 0x7f6d94ab5304 in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x27304) (BuildId: c047672cae7964324658491e7dee26748ae5d2f8)
    MariaDB#6 0x55c87aeef130 in _start (/build/sql/mariadbd+0x6ee130) (BuildId: abe0c6d0a0f0bdcee26c5e1f69353335cca3bd8a)

The sha1 has that is being encoded wasn't maked as initialized though
the openssl libs wrote the hash there.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Development

Successfully merging this pull request may close these issues.

5 participants