Bug 292692 - [exp-run] Update GCC_DEFAULT to 15
Summary: [exp-run] Update GCC_DEFAULT to 15
Status: In Progress
Alias: None
Product: Ports & Packages
Classification: Unclassified
Component: Individual Port(s) (show other bugs)
Version: Latest
Hardware: Any Any
: --- Affects Only Me
Assignee: FreeBSD Port Management Team (mail alias)
URL: https://gcc.gnu.org/gcc-15/porting_to...
Keywords:
Depends on: 293086 293282 293283 293284 293288 293290 293291 293294 293328 293331 293332 293333 293338 293342 293343 293087 293285 293286 293292 293308 293329 293330 293334 293339 293341 293364
Blocks:
  Show dependency tree / graph
 
Reported: 2026-01-24 11:07 UTC by Lorenzo Salvadore
Modified: 2026-07-17 06:49 UTC (History)
4 users (show)

See Also:
salvadore: exp-run?


Attachments
GCC_DEFAULT=15 (793 bytes, patch)
2026-01-24 11:07 UTC, Lorenzo Salvadore
no flags Details | Diff
kBuild gcc15 fix v1 (487 bytes, patch)
2026-02-01 09:52 UTC, Vladimir Druzenko
no flags Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Lorenzo Salvadore freebsd_committer freebsd_triage 2026-01-24 11:07:27 UTC
Created attachment 267432 [details]
GCC_DEFAULT=15

Let's start the process to update GCC_DEFAULT to 15. Patch is attached.
Comment 1 Michael Osipov freebsd_committer freebsd_triage 2026-01-24 18:54:27 UTC
(In reply to Lorenzo Salvadore from comment #0)

Why? 14 just settled in...
Comment 2 Lorenzo Salvadore freebsd_committer freebsd_triage 2026-01-25 11:06:55 UTC
GCC_DEFAULT updates always take many months. For example for GCC_DEFAULT=14 we started in August 2024 and finished in December 2025. So do not expect to have GCC_DEFAULT=15 in ports tree before spring, maybe even summer (but hopefully still in 2026!).

First GCC 15 release was in April 2025. I would say that we are not starting early to test GCC_DEFAULT=15, but rather that we committed late GCC_DEFAULT=14 (first release: May 2024): in addition to the usual dependent ports fixing/deprecating/updating we also had to wait for a FreeBSD release to go EoL this time, which increased the delay.
The first release of GCC 16 is already at stage 4 of development, so I think that it is likely that GCC 16.1 will be released before we finish with this exp-run, i.e. GCC_DEFAULT would still stay 1 version behind highest major upstream version (which is however enough to have GCC_DEFAULT on an upstream maintained release).
Comment 3 Michael Osipov freebsd_committer freebsd_triage 2026-01-25 19:58:52 UTC
I see, thank you for your explanation. I remember when I started to move from 13 to 15 with our inhouse C/Fortran code it took me a bit of time to fix it and some I haven't and added CLI options for the time being.
Comment 4 Antoine Brodin freebsd_committer freebsd_triage 2026-02-01 09:03:12 UTC
Some new failures logs:

https://pkg-status.freebsd.org/gohan04/data/143i386-default-foo/2026-01-30_10h01m31s/logs/errors/ircd-ratbox-3.0.10_8.log
https://pkg-status.freebsd.org/gohan04/data/143i386-default-foo/2026-01-30_10h01m31s/logs/errors/openblas-0.3.30,2.log

https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/alliance-5.1.1_6.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/freehdl-0.0.8_1.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/graywolf-0.1.6.4_4.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/proxysql-2.6.6_1.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/growl-for-linux-0.8.5_11.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/kBuild-0.1.9998_16.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/rhtvision-2.2.1.1_11.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/smv-2.5.4.3_11.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/xtensa-esp-elf-13.2.0.20240530_13.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/wine-proton-9.0.3.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/dxx-rebirth-20250329_1.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/xbat-1.11_11.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/rawtherapee-5.12.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/gforth-0.7.3_15.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/mono5.10-5.10.1.57_7.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/opencoarrays-2.10.1_1.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/gretl-2024d_2.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/octave-forge-mboct-fem-pkg-0.2.2.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/yate-6.4.1_1.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/q-6.0_1.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/grub2-pcbsd-2.02q_19.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/wiimms-2.23.a_10.log
https://pkg-status.freebsd.org/gohan05/data/143amd64-default-foo/2026-01-30_10h01m25s/logs/errors/kannel-1.4.4_15,1.log
Comment 5 Vladimir Druzenko freebsd_committer freebsd_triage 2026-02-01 09:52:07 UTC
Created attachment 267662 [details]
kBuild gcc15 fix v1

(In reply to Antoine Brodin from comment #4)
Can you please test this patch.
Comment 6 commit-hook freebsd_committer freebsd_triage 2026-02-02 18:28:06 UTC
A commit in branch main references this bug:

URL: https://cgit.FreeBSD.org/ports/commit/?id=614cec67767cb8afb47cfd290d17dbd812a7386e

commit 614cec67767cb8afb47cfd290d17dbd812a7386e
Author:     Vladimir Druzenko <vvd@FreeBSD.org>
AuthorDate: 2026-02-02 18:17:20 +0000
Commit:     Vladimir Druzenko <vvd@FreeBSD.org>
CommitDate: 2026-02-02 18:27:56 +0000

    devel/kBuild: Update 0.1.9998r3660 => 0.1.9998r3687, fix build with GCC 15

    Commit log:
    https://trac.netlabs.org/kbuild/changeset?reponame=&new=3687%40trunk&old=3660%40trunk

    PR:     292692 292801

 devel/kBuild/Makefile                              | 31 +++++++++-------------
 devel/kBuild/distinfo                              |  6 ++---
 devel/kBuild/files/patch-src_kmk_getopt.h (new)    | 11 ++++++++
 .../files/patch-src_kmk_glob_fnmatch.c (new)       | 11 ++++++++
 .../files/patch-src_oldsed_lib_getopt.h (new)      | 11 ++++++++
 devel/kBuild/pkg-plist                             |  2 ++
 6 files changed, 51 insertions(+), 21 deletions(-)
Comment 7 Lorenzo Salvadore freebsd_committer freebsd_triage 2026-02-21 16:09:40 UTC
All bug reports for this run have been created.
Comment 8 Matthias Andree freebsd_committer freebsd_triage 2026-02-22 13:49:01 UTC
Re the nine (9) complaints about "too many arguments" <https://bugs.freebsd.org/bugzilla/showdependencytree.cgi?id=292692&hide_resolved=1>:

NOTE: GCC15 changed the default C standard to build with to C23,
and the prominent change is that:

return_t functionname(); 

used to mean "unspecified arguments" from K&R (pre-C89) all the way through C17/C18

And C23 changed that to mean "exactly zero arguments".

The easy way out would be that the failing ports specify (in their Makefile) USE_CSTD=gnu17 or something similar to request the interpretation of such declarations according to older standard versions. See man gcc15 or man clang, look for -std= - valid ones: c89 c99 c11 c17 gnu89 gnu99 gnu11 gnu17 (c23 and gnu23 are valid in GCC15 but will probably fail, as reported by salvadore@).

BEWARE: GCC15 compile- and build-time checks through systems such as autoconf, cmake, meson, scons, and others, may change outcome and conclude that a certain function be unavailable.  

@Maintainers and committers on linked bug reports, be sure to carefully compare the "make configure" output with GCC 14 (even 13 if handy) and 15 and see where they change - and take that to the upstream package maintainers.
Comment 9 commit-hook freebsd_committer freebsd_triage 2026-04-26 12:44:30 UTC
A commit in branch main references this bug:

URL: https://cgit.FreeBSD.org/ports/commit/?id=b16813c9c2bcb5f8273a0788d35e88ed8394f47a

commit b16813c9c2bcb5f8273a0788d35e88ed8394f47a
Author:     Matthias Andree <mandree@FreeBSD.org>
AuthorDate: 2026-02-22 20:51:27 +0000
Commit:     Gleb Popov <arrowd@FreeBSD.org>
CommitDate: 2026-04-26 12:44:09 +0000

    graphics/rawtherapee: Add GCC (14) build option, clean up

    Clear the way for making GCC 15 the default GCC version for ports
    (through USE_GCC=yes), and set USE_GCC=14 for this port.

    GCC 15 and libc++ aren't compatible today because our (base) libc++'s
    header files around C++ type traits are rejected by GCC 15, which wants
    us to use library traits instead of (new) trait built-ins such as __decay
    on function signatures.  This isn't trivial to fix, so use the version
    pin to GCC 14 so that everyone else can benefit from 15.

    (Using libstdc++ with GCC instead would require the C++ libraries that
    expose C++ ABIs in the ports tree, Exiv2, Glib-derivatives be recompiled
    in a libstdc++ flavour and possibly be duplicated in a separate install
    path. Let's not go to these double efforts yet.)

    WARNING: on FreeBSD 15.0 amd64, clang-compiled rawtherapee requires
    twice as much CPU time on a certain test file batch conversion set than
    the GCC-compiled rawtherapee, so if you can, leave the GCC option on.

    While here, clean up the Makefile a bit.

    NOTE: the clang default build needs options tweaked a bit differently,
    and since the base compiler does not look into ${LOCALBASE}/include and
    ${LOCALBASE}/lib by default, we need to patch the rtengine/LibRaw.cmake
    wrapper to pass down proper -I and -L flags to libraw's
    autoconf/automake/libtool rig; we obtain them through pkg-config because
    cmake, although it already knows where to look, does not expose them
    through variables, as of 2026-02-22's ports cmake version.

    PR:             292692 (making GCC 15 default version)
    PR:             293329 (rawtherapee)
    Reported by:    salvadore@
    Pull Request:   https://github.com/freebsd/freebsd-ports/pull/504

 graphics/rawtherapee/Makefile                      | 86 +++++++++++-----------
 .../rawtherapee/files/patch-CMakeLists.txt (new)   | 13 ++++
 .../files/patch-rtengine_LibRaw.cmake (new)        | 30 ++++++++
 3 files changed, 88 insertions(+), 41 deletions(-)