Repository navigation
Closed
Labels
openssl-4-migrationRelated to switching to an OpenSSL 4 dependencyRelated to switching to an OpenSSL 4 dependency
Description
openssl@3 to openssl@4 migration tracking
-
ruby-build- ruby-build: migrate toopenssl@4#320610 -
xmrig- xmrig: migrate toopenssl@4#320711
Deferred with comment to revisit (9)
-
bareos-client(waiting for next release: matrix: add Fedora45 and fix OpenSSL 4 build errors in TLS/crypto code bareos/bareos#2797) -
ntp(upstream: Enable OpenSSL 4 builds while remaining OpenSSL 3 compatible ntp-project/ntp#32) -
prowler(blocked by Python < 3.14 requirement) -
s-nail(waiting for next release: sdaoden/s-mailx@e18fd33) -
sccache(upstream: sccache fails to build against OpenSSL 4.0 mozilla/sccache#2780) -
squid(upstream: Bug 5559: pt1: Fix bad uses of OpenSSL 0.9 API squid-cache/squid#2441, Bug 5559: pt2: Fix OpenSSL 4.0 compatibility squid-cache/squid#2501) -
ssldump(upstream: Add OpenSSL 4.0 support adulau/ssldump#168) -
sslscan(upstream: Build fails with OpenSSL 4 rbsec/sslscan#361) -
u-boot-tools(upstream: https://lists.u-boot-project.org/pipermail/u-boot/2026-April/616722.html)
Deferred until upstream support (10)
-
alpine(Debian patch: https://sources.debian.org/data/main/a/alpine/2.26+dfsg-6/debian/patches/openssl4.patch) -
mender-artifact- need support in https://github.com/mendersoftware/openssl (Ubuntu patch: https://bugs.launchpad.net/bugs/2154947) -
mender-cli- need support in https://github.com/mendersoftware/openssl (Ubuntu patch: https://bugs.launchpad.net/bugs/2154947) -
tt- need support in https://github.com/tarantool/go-openssl -
thors-anvil -
nagios-plugins -
nrpe(blocked bynagios-plugins) -
mysql-connector-c++ -
ldid -
ldid-procursus
Keep on openssl@3 (48)
20 old versions, e.g. .NET < 10, Node < 24, Python < 3.14, MySQL < 8.4, Ruby < 3.3
ansible@12- due topython@3.13dotnet@8dotnet@9haproxy@2.8node@20node@22pypy- Python 2 code basepypy3.11pypy3.12python@3.10python@3.11python@3.12python@3.13python@3.9tcl-tk@8- due topython-tk@3.11and oldermysql-client@8.0mysql@8.0percona-server@8.0percona-xtrabackup@8.0ruby@3.2
6 Deprecated
asuka- asuka: deprecate #318845authoscope- authoscope: deprecate #318846codeberg-cli- no conflicting OpenSSL versionsh2o- no conflicting OpenSSL versionsnoseyparker- no conflicting OpenSSL versionszurl- Build failure with OpenSSL 4.0 fanout/zurl#27
22 Disabled
ansible@10ansible@9conduredotnet@6ekg2git-serieshttp_loadlibgit2@1.7mariadb@10.5node@18omnaraorogenepostgresql@12postgresql@13pypy3.10redis@6.2rethinkdbruby@3.1ryescreenpipesgrvineyard
https://openssl-library.org/post/2026-04-14-openssl-40-final-release/) for breaking changes. Common failure modes:
- C/C++:
ASN1_STRINGis now opaque; X509 getters now returnconstpointers. - Rust:
openssl-sys v0.10.78(released 2026-04-19) is the first release supporting OpenSSL 4. - Some formulae declare
depends_on "openssl@3"but do not actually link it. Verify if we can drop the dependency rather than migrate.
Related:
Metadata
Metadata
Assignees
Labels
openssl-4-migrationRelated to switching to an OpenSSL 4 dependencyRelated to switching to an OpenSSL 4 dependency
Activity
GunniBusch commentedon Apr 26, 2026
One question, what's the current policy regarding formulae using openssl@3 but recursive dependencies using openssl@4 or vice versa?
Because let's take
cryptography, which would supportopenssl@4but it has a very huge amount of dependents that useopenss@3somehow.Technically the only solution without policy change would to have a pr that updates all conflicting formulae at once, which would result in some prs with like 300 changed formulae, and also unnecessarily keeping formulae at an older version of openssl
chenrui333 commentedon May 4, 2026
I am starting the staging branch setup now
GunniBusch commentedon May 4, 2026
noice.
GunniBusch commentedon May 5, 2026
Sorry for bothering you, but since we go root first (as mentioned in one of my prs by you), would it be preferred to have one pr per formula or like when we removed no auto bump, multiple like 20 or so formulae per pr? I am not fully understanding when it is fine for having those a* prs and when it is preferred to have single formula prs.
p-linnane commentedon May 5, 2026
@GunniBusch We tend to do singles unless there's a compelling reason. Odds of something going wrong are higher with a multi-hour PR of a ton of formulae.
GunniBusch commentedon May 5, 2026
Makes sense. So when we only have syntax change like with autobump, there was not really something that could fail/break. Got it
p-linnane commentedon May 5, 2026
Yes, syntax only PR's don't go through the entire build process, so they're quick.
ferdymercury commentedon Sep 17, 2026
For civetweb formula, this patch will be needed: civetweb/civetweb#1423
baeeica7 commentedon Oct 15, 2026
cho-m commentedon Oct 17, 2026
Closing as majority of migration is done. Remaining is waiting on upstream and can be tracked per formula.
p-linnane commentedon Oct 17, 2026
Thanks for all the great work here @cho-m!