Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.kernel > #90602 > unrolled thread

s390x: removal of 31-bit libraries

Back to article view | Back to linux.debian.kernel

Contents

  s390x: removal of 31-bit libraries Aurelien Jarno <aurel32@debian.org> - 2025-12-29 13:20 +0100
    Re: s390x: removal of 31-bit libraries Sebastian Ramacher <sramacher@debian.org> - 2026-04-28 10:30 +0200
      Re: s390x: removal of 31-bit libraries Aurelien Jarno <aurel32@debian.org> - 2026-04-28 18:50 +0200
        Re: s390x: removal of 31-bit libraries Aurelien Jarno <aurel32@debian.org> - 2026-06-05 20:20 +0200
          Re: s390x: removal of 31-bit libraries Aurelien Jarno <aurel32@debian.org> - 2026-06-10 06:40 +0200
            Re: s390x: removal of 31-bit libraries Sebastian Ramacher <sramacher@debian.org> - 2026-06-10 08:30 +0200

#90602 — s390x: removal of 31-bit libraries

[Multipart message — attachments visible in raw view] — view raw

Hi,

Linux 6.19 will drop support for 31-bit compatibility on s390x, and will 
also stop providing 31-bit support in UAPI header files [1]. Once this 
lands in unstable, it will become impossible to build the libc6-s390 and 
libc6-dev-s390 packages on s390x. We should therefore stop building 
these packages, as well as g++-15-multilib and the corresponding lib32* 
libraries.

Doing so will break many packages. I therefore suggest that, over the 
next few weeks, we start removing 31-bit support in all packages except 
src:gcc-15, src:gcc-defaults, src:glibc, src:linux and src:zlib (and 
possibly the corresponding cross compilers?). Then in a final 
coordinated step, we can fully drop 31-bit support from those remaining 
packages as well.

Any thoughts about that? Volunteers to coordinate this? 

Regards
Aurelien

[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=3db5cf9354710c3b9268af74a0dca88514279453

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                     http://aurel32.net

[toc] | [next] | [standalone]

#92146

Hi

On 2025-12-29 13:12:48 +0100, Aurelien Jarno wrote:
> Hi,
> 
> Linux 6.19 will drop support for 31-bit compatibility on s390x, and will 
> also stop providing 31-bit support in UAPI header files [1]. Once this 
> lands in unstable, it will become impossible to build the libc6-s390 and 
> libc6-dev-s390 packages on s390x. We should therefore stop building 
> these packages, as well as g++-15-multilib and the corresponding lib32* 
> libraries.
> 
> Doing so will break many packages. I therefore suggest that, over the 
> next few weeks, we start removing 31-bit support in all packages except 
> src:gcc-15, src:gcc-defaults, src:glibc, src:linux and src:zlib (and 
> possibly the corresponding cross compilers?). Then in a final 
> coordinated step, we can fully drop 31-bit support from those remaining 
> packages as well.
> 
> Any thoughts about that? Volunteers to coordinate this? 

If I am reading dak output correctly, we are down to gcc-11, gcc-13,
gcc-14, gcc-15, zlib, and glibc itself. So assuming that I didn't miss
anything, are we there to also drop the support from these packages?

Cheers
-- 
Sebastian Ramacher

[toc] | [prev] | [next] | [standalone]

#92152

Hi,

On 2026-04-28 10:22, Sebastian Ramacher wrote:
> Hi
> 
> On 2025-12-29 13:12:48 +0100, Aurelien Jarno wrote:
> > Hi,
> > 
> > Linux 6.19 will drop support for 31-bit compatibility on s390x, and will 
> > also stop providing 31-bit support in UAPI header files [1]. Once this 
> > lands in unstable, it will become impossible to build the libc6-s390 and 
> > libc6-dev-s390 packages on s390x. We should therefore stop building 
> > these packages, as well as g++-15-multilib and the corresponding lib32* 
> > libraries.
> > 
> > Doing so will break many packages. I therefore suggest that, over the 
> > next few weeks, we start removing 31-bit support in all packages except 
> > src:gcc-15, src:gcc-defaults, src:glibc, src:linux and src:zlib (and 
> > possibly the corresponding cross compilers?). Then in a final 
> > coordinated step, we can fully drop 31-bit support from those remaining 
> > packages as well.
> > 
> > Any thoughts about that? Volunteers to coordinate this? 
> 
> If I am reading dak output correctly, we are down to gcc-11, gcc-13,
> gcc-14, gcc-15, zlib, and glibc itself. So assuming that I didn't miss
> anything, are we there to also drop the support from these packages?

These packages have already been fixed, but they are kept as cruft in 
sid (but not in forky). I have not investigated why, it could just be 
that dak is not smart enough.

In addition llvm-toolchain-{14,19,20} are still waiting for a fix [1], 
but they are only in sid or experimental.

Regards
Aurelien

[1] https://bugs.debian.org/cgi-bin/pkgreport.cgi?tag=31-bit-removal;users=debian-s390@lists.debian.org

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                     http://aurel32.net

[toc] | [prev] | [next] | [standalone]

#92906

Hi,

A small status update.

On 2026-04-28 18:47, Aurelien Jarno wrote:
> Hi,
> 
> On 2026-04-28 10:22, Sebastian Ramacher wrote:
> > Hi
> > 
> > On 2025-12-29 13:12:48 +0100, Aurelien Jarno wrote:
> > > Hi,
> > > 
> > > Linux 6.19 will drop support for 31-bit compatibility on s390x, and will 
> > > also stop providing 31-bit support in UAPI header files [1]. Once this 
> > > lands in unstable, it will become impossible to build the libc6-s390 and 
> > > libc6-dev-s390 packages on s390x. We should therefore stop building 
> > > these packages, as well as g++-15-multilib and the corresponding lib32* 
> > > libraries.
> > > 
> > > Doing so will break many packages. I therefore suggest that, over the 
> > > next few weeks, we start removing 31-bit support in all packages except 
> > > src:gcc-15, src:gcc-defaults, src:glibc, src:linux and src:zlib (and 
> > > possibly the corresponding cross compilers?). Then in a final 
> > > coordinated step, we can fully drop 31-bit support from those remaining 
> > > packages as well.
> > > 
> > > Any thoughts about that? Volunteers to coordinate this? 
> > 
> > If I am reading dak output correctly, we are down to gcc-11, gcc-13,
> > gcc-14, gcc-15, zlib, and glibc itself. So assuming that I didn't miss
> > anything, are we there to also drop the support from these packages?
> 
> These packages have already been fixed, but they are kept as cruft in 
> sid (but not in forky). I have not investigated why, it could just be 
> that dak is not smart enough.

All the cross packages got removed. gcc-1[1-4] also got removed.

> In addition llvm-toolchain-{14,19,20} are still waiting for a fix [1], 
> but they are only in sid or experimental.

llvm-toolchain-14 is also gone, for all architectures.

For dependencies, we are left with only gcc-15, gcc-defaults, zlib and 
glibc, I'll work on filling the corresponding bugs against 
ftp.debian.org as the next step.

For build-dependencies, llvm-toolchain-{19,20} still need to be fixed.

Regards
Aurelien

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                     http://aurel32.net

[toc] | [prev] | [next] | [standalone]

#92938

Hi,

On 2026-06-05 20:13, Aurelien Jarno wrote:
> Hi,
> 
> A small status update.
> 
> On 2026-04-28 18:47, Aurelien Jarno wrote:
> > Hi,
> > 
> > On 2026-04-28 10:22, Sebastian Ramacher wrote:
> > > Hi
> > > 
> > > On 2025-12-29 13:12:48 +0100, Aurelien Jarno wrote:
> > > > Hi,
> > > > 
> > > > Linux 6.19 will drop support for 31-bit compatibility on s390x, and will 
> > > > also stop providing 31-bit support in UAPI header files [1]. Once this 
> > > > lands in unstable, it will become impossible to build the libc6-s390 and 
> > > > libc6-dev-s390 packages on s390x. We should therefore stop building 
> > > > these packages, as well as g++-15-multilib and the corresponding lib32* 
> > > > libraries.
> > > > 
> > > > Doing so will break many packages. I therefore suggest that, over the 
> > > > next few weeks, we start removing 31-bit support in all packages except 
> > > > src:gcc-15, src:gcc-defaults, src:glibc, src:linux and src:zlib (and 
> > > > possibly the corresponding cross compilers?). Then in a final 
> > > > coordinated step, we can fully drop 31-bit support from those remaining 
> > > > packages as well.
> > > > 
> > > > Any thoughts about that? Volunteers to coordinate this? 
> > > 
> > > If I am reading dak output correctly, we are down to gcc-11, gcc-13,
> > > gcc-14, gcc-15, zlib, and glibc itself. So assuming that I didn't miss
> > > anything, are we there to also drop the support from these packages?
> > 
> > These packages have already been fixed, but they are kept as cruft in 
> > sid (but not in forky). I have not investigated why, it could just be 
> > that dak is not smart enough.
> 
> All the cross packages got removed. gcc-1[1-4] also got removed.
> 
> > In addition llvm-toolchain-{14,19,20} are still waiting for a fix [1], 
> > but they are only in sid or experimental.
> 
> llvm-toolchain-14 is also gone, for all architectures.
> 
> For dependencies, we are left with only gcc-15, gcc-defaults, zlib and 
> glibc, I'll work on filling the corresponding bugs against 
> ftp.debian.org as the next step.

This is now done. A bit thanks to Thorsten Alteholz for handling all the 
removal request. This means we do not have 31-bit s390 binaries anymore 
in testing or unstable \o/.

Release team members, I guess that with done, the "s390-31-bit-rm" 
transition can be remove from the tracker.

> For build-dependencies, llvm-toolchain-{19,20} still need to be fixed.

This part still needs to be fixed but requires new sourceful uploads.  
There are RC bugs (#1125251 and #1125252) for tracking that.

In addition both llvm-toolchain-19 and llvm-toolchain-20 currently FTBFS 
due to CMake 4 (#1113237 and #1139588). In addition llvm-toolchain-20 is 
only kept in experimental to possibly help the removal of older 
llvm-toolchain versions (see #1116216) if some packages can't switch to 
llvm-toolchain-21.

I am therefore not planning to work on fixing these packages.

Regards
Aurelien

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                     http://aurel32.net

[toc] | [prev] | [next] | [standalone]

#92939

On 2026-06-10 06:38:03 +0200, Aurelien Jarno wrote:
> Hi,
> 
> On 2026-06-05 20:13, Aurelien Jarno wrote:
> > Hi,
> > 
> > A small status update.
> > 
> > On 2026-04-28 18:47, Aurelien Jarno wrote:
> > > Hi,
> > > 
> > > On 2026-04-28 10:22, Sebastian Ramacher wrote:
> > > > Hi
> > > > 
> > > > On 2025-12-29 13:12:48 +0100, Aurelien Jarno wrote:
> > > > > Hi,
> > > > > 
> > > > > Linux 6.19 will drop support for 31-bit compatibility on s390x, and will 
> > > > > also stop providing 31-bit support in UAPI header files [1]. Once this 
> > > > > lands in unstable, it will become impossible to build the libc6-s390 and 
> > > > > libc6-dev-s390 packages on s390x. We should therefore stop building 
> > > > > these packages, as well as g++-15-multilib and the corresponding lib32* 
> > > > > libraries.
> > > > > 
> > > > > Doing so will break many packages. I therefore suggest that, over the 
> > > > > next few weeks, we start removing 31-bit support in all packages except 
> > > > > src:gcc-15, src:gcc-defaults, src:glibc, src:linux and src:zlib (and 
> > > > > possibly the corresponding cross compilers?). Then in a final 
> > > > > coordinated step, we can fully drop 31-bit support from those remaining 
> > > > > packages as well.
> > > > > 
> > > > > Any thoughts about that? Volunteers to coordinate this? 
> > > > 
> > > > If I am reading dak output correctly, we are down to gcc-11, gcc-13,
> > > > gcc-14, gcc-15, zlib, and glibc itself. So assuming that I didn't miss
> > > > anything, are we there to also drop the support from these packages?
> > > 
> > > These packages have already been fixed, but they are kept as cruft in 
> > > sid (but not in forky). I have not investigated why, it could just be 
> > > that dak is not smart enough.
> > 
> > All the cross packages got removed. gcc-1[1-4] also got removed.
> > 
> > > In addition llvm-toolchain-{14,19,20} are still waiting for a fix [1], 
> > > but they are only in sid or experimental.
> > 
> > llvm-toolchain-14 is also gone, for all architectures.
> > 
> > For dependencies, we are left with only gcc-15, gcc-defaults, zlib and 
> > glibc, I'll work on filling the corresponding bugs against 
> > ftp.debian.org as the next step.
> 
> This is now done. A bit thanks to Thorsten Alteholz for handling all the 
> removal request. This means we do not have 31-bit s390 binaries anymore 
> in testing or unstable \o/.
> 
> Release team members, I guess that with done, the "s390-31-bit-rm" 
> transition can be remove from the tracker.

Thanks for driving this removal. I have removed the trackers.

Cheers
-- 
Sebastian Ramacher

[toc] | [prev] | [standalone]

Back to top | Article view | linux.debian.kernel

csiph-web