Discussion:
Bug#1081553: transition: abseil
Add Reply
Benjamin Barenblat
5 months ago
Reply
Permalink
Package: release.debian.org
Severity: normal
User: ***@packages.debian.org
Usertags: transition
X-Debbugs-Cc: ***@packages.debian.org, Stefano Rivera <***@debian.org>
Control: affects -1 + src:abseil

I'd like to transition sid from Abseil 20230802 to Abseil 20240722. The new
version has a new ABI (with a new SONAME and new binary package names).

Tests for 20240722.0-1 in experimental are green on all supported architectures.

A number of packages in sid depend directly on Abseil. To give early warning of
breakages, I've done trial rebuilds as appropriate on the amd64 porterbox.
Packages that work fine with the new version:

- dm-tree
- libgav1
- libphonenumber
- mozc
- mujoco
- open-vm-tools
- protobuf
- re2
- ycmd

Packages that are broken by the new version:

- falcosecurity-libs: FTBFS because it depends both on Abseil directly
and on Abseil via RE2, and the RE2 in unstable hasn't been built
against the new Abseil.

- grpc: FTBFS because it depends both on Abseil directly and on Abseil
via Protobuf, and the Protobuf in unstable hasn't been built against
the new Abseil.

- s2geometry: FTBFS because it hard-codes std=c+11 and the new version
requires -std=c++14 or later (https://bugs.debian.org/1059228)

- webrtc-audio-processing: FTBFS because it relies on transitive
#includes that have changed

Packages that I'm not sure about:

- firebird4.0: has an active FTBFS (https://bugs.debian.org/1079523)

- ortools: has an active FTBFS (https://bugs.debian.org/1024790)

- libreoffice: too big to build on a porterbox, so left untested

None of the breakages in sid should be challenging to repair.

Ben file:

title = "abseil";
is_affected = .depends ~ /\blibabsl20230802\b/ | .depends ~ /\blibabsl20240722\b/ | ;
is_good = .depends ~ /\blibabsl20240722\b/;
is_bad = .depends ~ /\blibabsl20230802\b/;
Debian Bug Tracking System
5 months ago
Reply
Permalink
tags -1 moreinfo
Bug #1081553 [release.debian.org] transition: abseil
Added tag(s) moreinfo.
forwarded -1 https://release.debian.org/transitions/html/auto-abseil.html
Bug #1081553 [release.debian.org] transition: abseil
Set Bug forwarded-to-address to 'https://release.debian.org/transitions/html/auto-abseil.html'.
--
1081553: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1081553
Debian Bug Tracking System
Contact ***@bugs.debian.org with problems
Sebastian Ramacher
5 months ago
Reply
Permalink
Control: tags -1 moreinfo
Control: forwarded -1 https://release.debian.org/transitions/html/auto-abseil.html
...
Thanks for doing the test builds. Do you also have test results for
llvm-toolchain-{15,16,17,18,19}?

Cheers
--
Sebastian Ramacher
Benjamin Barenblat
5 months ago
Reply
Permalink
Do you also have test results for llvm-toolchain-{15,16,17,18,19}?
I haven't actually tried them. They don't depend on Abseil directly,
just on gRPC. I thus expect they will be broken by the transition until
gRPC gets binNMU'd. Let me know if you'd like me to try rebuilding them
anyway.

Benjamin
Sebastian Ramacher
5 months ago
Reply
Permalink
Post by Benjamin Barenblat
Do you also have test results for llvm-toolchain-{15,16,17,18,19}?
I haven't actually tried them. They don't depend on Abseil directly,
just on gRPC. I thus expect they will be broken by the transition until
gRPC gets binNMU'd. Let me know if you'd like me to try rebuilding them
anyway.
As they are key packages it would be good to know if we can rebuild them
without issues. If a rebuilt grpc would help to do the tests, I can
schedule binNMUs in experimental to have grpc built against the new
abseil.

Cheers
--
Sebastian Ramacher
Benjamin Barenblat
4 months ago
Reply
Permalink
If a rebuilt grpc would help to do the tests, I can schedule binNMUs in
experimental to have grpc built against the new abseil.
BinNMUs for grpc in experimental would indeed be helpful. If you binNMU
gRPC, I can do llvm-toolchain rebuilds on one of the porterboxes.

Benjamin
Sebastian Ramacher
4 months ago
Reply
Permalink
Post by Benjamin Barenblat
If a rebuilt grpc would help to do the tests, I can schedule binNMUs in
experimental to have grpc built against the new abseil.
BinNMUs for grpc in experimental would indeed be helpful. If you binNMU
gRPC, I can do llvm-toolchain rebuilds on one of the porterboxes.
binNMU scheduled for grpc in experimental.

Cheers
--
Sebastian Ramacher
Sebastian Ramacher
2 months ago
Reply
Permalink
Hi
Post by Sebastian Ramacher
Post by Benjamin Barenblat
If a rebuilt grpc would help to do the tests, I can schedule binNMUs in
experimental to have grpc built against the new abseil.
BinNMUs for grpc in experimental would indeed be helpful. If you binNMU
gRPC, I can do llvm-toolchain rebuilds on one of the porterboxes.
binNMU scheduled for grpc in experimental.
Were the test rebuilds for llvm-toolchain-* successful?

Cheers
--
Sebastian Ramacher
Benjamin Barenblat
2 months ago
Reply
Permalink
Post by Sebastian Ramacher
Were the test rebuilds for llvm-toolchain-* successful?
I haven’t tried rebuilding yet. I’ll see if I can get to it this week.

Benjamin
Benjamin Barenblat
about a month ago
Reply
Permalink
Post by Sebastian Ramacher
Were the test rebuilds for llvm-toolchain-* successful?
llvm-toolchain-17, -18, and -19 build successfully on amd64 with the new
Abseil packages installed
(sbuild --extra-package=.../libabsl20240722_20240722.0-1_amd64.deb
--extra-package=.../libabsl-dev_20240722.0-1_amd64.deb). -15 and -16 do
not build (-15 due to unresolvable dependencies and -16 due to an actual
build failure), but neither of these is in sid anymore, so it might be
okay.

Would you like me to do some builds on non-amd64 architectures as well?
Sebastian Ramacher
about a month ago
Reply
Permalink
...
Thanks for the tests! There is no need to also test it on non-amd64
architectures. We will start this transition once the current Python
3.11 as default transition is done.

Cheers
--
Sebastian Ramacher
Debian Bug Tracking System
5 months ago
Reply
Permalink
Post by Benjamin Barenblat
affects -1 + src:abseil
Bug #1081553 [release.debian.org] transition: abseil
Added indication that 1081553 affects src:abseil
--
1081553: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1081553
Debian Bug Tracking System
Contact ***@bugs.debian.org with problems
Rene Engelhard
4 months ago
Reply
Permalink
Hi,
Post by Benjamin Barenblat
- libreoffice: too big to build on a porterbox, so left untested
This is not the first time I see this claim.

FWIW, I don't buy this.

a) you only need -B

b) /dev/mapper/vg0-srv  159G     48G  105G   32% /srv
   and
   $ cat /proc/cpuinfo | grep ^cpu | wc -l
16
   and
   $ cat /proc/meminfo
MemTotal:        8131284 kB

is definitely more than suitable.


Test it. I do need to build big stuff too if I have a transition involving that and as we see above you perfectly can do this on a porterbox.

(Even with a full -b build)


Regards,


Rene
Stefano Rivera
21 days ago
Reply
Permalink
Hi Emilio (2025.01.26_11:43:57_-0400)
Also this update seems to cause a regression in re2, that needs to be
There are new upstream releases of re2 that deal with that. I've just
uploaded that to experimental.

Stefano
--
Stefano Rivera
http://tumbleweed.org.za/
+1 415 683 3272
Loading...