You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@thrift.apache.org by Jens Geyer <je...@hotmail.com> on 2025/11/16 11:48:54 UTC

Next Release

Hi all,

it is that time of the year again and so I would start putting a 0.23.0 
RC together somewhere in the next two weeks.

If there are any objections, e.g. because we want to fix/integrate 
something first that should make it into that release, just let me/us 
know. The date is not set in concrete, it would simply be nice to know 
early.

Have fun,
JensG




Re: Next Release

Posted by Carel Combrink <ca...@gmail.com>.
> In cmake, this could be combined with WINDOWS_EXPORT_ALL_SYMBOLS (see
>>
>> https://cmake.org/cmake/help/latest/prop_tgt/WINDOWS_EXPORT_ALL_SYMBOLS.html
>> ).
>>
>
> I would probably vote for `WINDOWS_EXPORT_ALL_SYMBOLS ` since it does not
> seem like Windows has a very large user base and to touch all the files for
> a smaller user base does not look productive, and it then has at least the
> same behaviour on Linux and Windows.
>

I see this is already set in cmake:
https://github.com/apache/thrift/blob/master/build/cmake/DefinePlatformSpecifc.cmake#L59

Re: Next Release

Posted by Mario Emmenlauer <ma...@emmenlauer.de.INVALID>.
Hi Carel,

On 11/17/25 12:33, Carel Combrink wrote:
> Hi Mario,
> 
>     tests on my local branch successfully. However, the PR is not sufficient
>     for a shared library build, at least not on MSVC, because the library
>     does not export any symbols. Maybe I'm going into old news here?
> 
> 
> This is curious since the the CI tests do not show this issue nor did the previous iterations of the conan package.
> If I understand correctly if none were exported then at least the conan packages should fail.
> The tests here are also expected to fail but I can think of many reasons why it might work.
> I will see if I can't find anything.
> PS: Conan has a very basic test package that I am lead to believe does work with previous versions: https://github.com/conan-io/conan-center-index/blob/master/ 
> recipes/thrift/all/test_package/test_package.cpp <https://github.com/conan-io/conan-center-index/blob/master/recipes/thrift/all/test_package/test_package.cpp>

Ok, I finally got around to checking this, and actually, the cmake build does
set `CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS` in `build/cmake/DefinePlatformSpecifc.cmake`.
This is then sufficient - but it will be cmake-only. There is a complex logic
behind this flag, that is not easily reproduced with autotools.


>     In cmake, this could be combined with WINDOWS_EXPORT_ALL_SYMBOLS (see
>     https://cmake.org/cmake/help/latest/prop_tgt/WINDOWS_EXPORT_ALL_SYMBOLS.html <https://cmake.org/cmake/help/latest/prop_tgt/WINDOWS_EXPORT_ALL_SYMBOLS.html>).
> 
> I would probably vote for `WINDOWS_EXPORT_ALL_SYMBOLS ` since it does not seem like Windows has a very large user base and to touch all the files for a smaller 
> user base does not look productive, and it then has at least the same behaviour on Linux and Windows.
> 
>     To continue automake support, one could also combine it with "real" export
>     macros like for example https://wiki.tcl-lang.org/page/DLLEXPORT+and+DLLIMPORT <https://wiki.tcl-lang.org/page/DLLEXPORT+and+DLLIMPORT>
> 
> cmake has the ability to generate the import/export macros but thrift is not built only by cmake which makes it less ideal.
> See https://cmake.org/cmake/help/latest/module/GenerateExportHeader.html <https://cmake.org/cmake/help/latest/module/GenerateExportHeader.html>

Yes, the import/export macros would be the cleaner way. What cmake generates
is actually easily reproduced in autotools, it just boils down to a new
header, and a few defines. And thrift contains such a header already in
thrift/thrift_export.h, however it is currently not (or sparsely) used.


>     Did you consider this?
> 
> No, as mentioned above
> 
> Thank you for taking the time to look at this,
> Regards,
> Carel



Viele Grüße,

     Mario Emmenlauer


Re: Next Release

Posted by Carel Combrink <ca...@gmail.com>.
Hi Mario,


> tests on my local branch successfully. However, the PR is not sufficient
> for a shared library build, at least not on MSVC, because the library
> does not export any symbols. Maybe I'm going into old news here?
>

This is curious since the the CI tests do not show this issue nor did the
previous iterations of the conan package.
If I understand correctly if none were exported then at least the conan
packages should fail.
The tests here are also expected to fail but I can think of many reasons
why it might work.
I will see if I can't find anything.
PS: Conan has a very basic test package that I am lead to believe does work
with previous versions:
https://github.com/conan-io/conan-center-index/blob/master/recipes/thrift/all/test_package/test_package.cpp


> In cmake, this could be combined with WINDOWS_EXPORT_ALL_SYMBOLS (see
>
> https://cmake.org/cmake/help/latest/prop_tgt/WINDOWS_EXPORT_ALL_SYMBOLS.html
> ).
>

I would probably vote for `WINDOWS_EXPORT_ALL_SYMBOLS ` since it does not
seem like Windows has a very large user base and to touch all the files for
a smaller user base does not look productive, and it then has at least the
same behaviour on Linux and Windows.


> To continue automake support, one could also combine it with "real" export
> macros like for example
> https://wiki.tcl-lang.org/page/DLLEXPORT+and+DLLIMPORT


cmake has the ability to generate the import/export macros but thrift is
not built only by cmake which makes it less ideal.
See https://cmake.org/cmake/help/latest/module/GenerateExportHeader.html


> Did you consider this?
>
No, as mentioned above

Thank you for taking the time to look at this,
Regards,
Carel

Re: Next Release

Posted by Mario Emmenlauer <ma...@emmenlauer.de.INVALID>.
Dear Carel,

On 11/17/25 06:59, Carel Combrink wrote:
> And on the C++ side would really like to get this one done: "Unable to
> build Thrift as a shared library on Windows"
> https://issues.apache.org/jira/browse/THRIFT-5898
Your PR about GlobalOutput looks good to me, and I recently made similar
tests on my local branch successfully. However, the PR is not sufficient
for a shared library build, at least not on MSVC, because the library
does not export any symbols. Maybe I'm going into old news here?

In cmake, this could be combined with WINDOWS_EXPORT_ALL_SYMBOLS (see
https://cmake.org/cmake/help/latest/prop_tgt/WINDOWS_EXPORT_ALL_SYMBOLS.html).
To continue automake support, one could also combine it with "real" export
macros like for example https://wiki.tcl-lang.org/page/DLLEXPORT+and+DLLIMPORT

Did you consider this?

All the best,

     Mario



Re: Next Release

Posted by Carel Combrink <ca...@gmail.com>.
Hi  JensG,

I have the following on the Python side I would like to have pushed over
the line:
- Python housekeeping: https://github.com/apache/thrift/pull/3231
- THRIFT-5885: Python incorrectly deserializes Enum:
https://github.com/apache/thrift/pull/3232
- Thrift-5900: Fix Tests for Python 3.14:
https://github.com/apache/thrift/pull/3239

And on the C++ side would really like to get this one done: "Unable to
build Thrift as a shared library on Windows"
https://issues.apache.org/jira/browse/THRIFT-5898
- I have an open PR https://github.com/apache/thrift/pull/3226 that needs
some eyes and consideration.
- It is a breaking change and I am not sure how many people use the
`GlobalOutput` feature of the thrift libs. For that reason I am
reconsidering the fix to rather try and add the above but in a backwards
compatible way and then deprecating the old way for people to at least have
one release to migrate code.
A solution to this I think must be implemented for the next release, even
if it is not perfect
Any feedback will be much appreciated.

Regards,
Carel



On Sun, 16 Nov 2025 at 13:52, Jens Geyer <je...@hotmail.com> wrote:

> Hi all,
>
> it is that time of the year again and so I would start putting a 0.23.0
> RC together somewhere in the next two weeks.
>
> If there are any objections, e.g. because we want to fix/integrate
> something first that should make it into that release, just let me/us
> know. The date is not set in concrete, it would simply be nice to know
> early.
>
> Have fun,
> JensG
>
>
>
>

Re: Next Release

Posted by Carel Combrink <ca...@gmail.com>.
Hi Jens,

I have one PR open on nodejs that I would like to get in, although I won't
be able to work on it for the next 7 days (holiday) if it has issues.

But please don't let this hold up the release, at least from my side.

Thanks for the effort and good luck,
Carel

On Fri, 27 Mar 2026, 13:02 Jens Geyer, <je...@hotmail.com> wrote:

> Hi,
>
> I would reactivate that thread again. Depending on the time slots
> available, I will create a release branch in the next days or so., after
> my two pending PRs got merged. If there are any blockers to a release
> 0.23.0, this is right the time to raise your hand.
>
> Have fun,
> JensG
>
>
> Am 16.11.2025 um 12:48 schrieb Jens Geyer:
> > Hi all,
> >
> > it is that time of the year again and so I would start putting a
> > 0.23.0 RC together somewhere in the next two weeks.
> >
> > If there are any objections, e.g. because we want to fix/integrate
> > something first that should make it into that release, just let me/us
> > know. The date is not set in concrete, it would simply be nice to know
> > early.
> >
> > Have fun,
> > JensG
> >
> >
> >
>

Re: Next Release

Posted by Carel Combrink <ca...@gmail.com>.
Hi Jens,

The quick-fix would be to add `--break-system-packages` to the `python3 -m
pip install` command to overcome the issue.
There might be a better long term solution that I am yet to find. My
opinion is that since it is in a docker container it should be fine to
suppress the warning using the flag, I might be wrong or mistaken.

PEP 668 explains this: https://peps.python.org/pep-0668/
Section 5 states: "A distro Python when used in a single-application
container image (e.g., a Docker container). In this use case, the risk of
breaking system software is lower, since generally only a single
application runs in the container, and the impact is lower, since you can
rebuild the container and you don’t have to struggle to recover a running
machine. There are also a large number of existing Dockerfiles with an
unqualified RUN pip install ... statement, etc., and it would be good not
to break those. So, builders of base container images may want to ensure
that the marker file is not present, even if the underlying OS ships one by
default."

I made a PR in this effect: https://github.com/apache/thrift/pull/3380

Regards,
Carel

On Wed, 8 Apr 2026 at 01:02, Jens Geyer <je...@hotmail.com> wrote:

> Hi,
>
> actually I could indeed need some help: This is what
>
>      docker build -t thrift build/docker/ubuntu-noble
>
> produces. Something wth the recent py changes is not perfect:
>
>
>   => ERROR [28/37] RUN python3 -m pip install --no-cache-dir --upgrade
> "tornado>=6.3.0" "twisted>=24.3.0" "zope.in  0.8s
> ------
>   > [28/37] RUN python3 -m pip install --no-cache-dir --upgrade
> "tornado>=6.3.0" "twisted>=24.3.0" "zope.interface>=6.1":
> 0.689 error: externally-managed-environment
> 0.689
> 0.689 × This environment is externally managed
> 0.689 ╰─> To install Python packages system-wide, try apt install
> 0.689     python3-xyz, where xyz is the package you are trying to
> 0.689     install.
> 0.689
> 0.689     If you wish to install a non-Debian-packaged Python package,
> 0.689     create a virtual environment using python3 -m venv path/to/venv.
> 0.689     Then use path/to/venv/bin/python and path/to/venv/bin/pip. Make
> 0.689     sure you have python3-full installed.
> 0.689
> 0.689     If you wish to install a non-Debian packaged Python application,
> 0.689     it may be easiest to use pipx install xyz, which will manage a
> 0.689     virtual environment for you. Make sure you have pipx installed.
> 0.689
> 0.689     See /usr/share/doc/python3.12/README.venv for more information.
> 0.689
> 0.689 note: If you believe this is a mistake, please contact your Python
> installation or OS distribution provider. You can override this, at the
> risk of breaking your Python installation or OS, by passing
> --break-system-packages.
> 0.689 hint: See PEP 668 for the detailed specification.
> ------
>
>   15 warnings found (use docker --debug to expand):
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 148)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 309)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 78)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 85)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 110)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 177)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 269)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 156)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 157)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 284)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 97)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 126)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 155)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 109)
>   - LegacyKeyValueFormat: "ENV key=value" should be used instead of
> legacy "ENV key value" format (line 130)
> Dockerfile:258
> --------------------
>   256 |           python3-wheel
>   257 |
>   258 | >>> RUN python3 -m pip install --no-cache-dir --upgrade
> "tornado>=6.3.0" "twisted>=24.3.0" "zope.interface>=6.1"
>   259 |
>   260 |     RUN apt-get install -y --no-install-recommends \
> --------------------
> ERROR: failed to build: failed to solve: process "/bin/sh -c python3 -m
> pip install --no-cache-dir --upgrade \"tornado>=6.3.0\"
> \"twisted>=24.3.0\" \"zope.interface>=6.1\"" did not complete
> successfully: exit code: 1
>
> The env key stuff can be ignored for now. But the error is a blocker.
>
> Thanks for any help,
> JensG
>
>
>
> Am 07.04.2026 um 14:20 schrieb Carel Combrink:
> > Hi Jens,
> >
> > I see my last PR was merged, thanks for that.
> >
> > Any progress on the release?
> >
> > Please shout if there is something that I can do to help if there is
> > something that needs a hand.
> >
> > Regards,
> > Carel
>

Re: Next Release

Posted by Jens Geyer <je...@hotmail.com>.
Hi,

actually I could indeed need some help: This is what

     docker build -t thrift build/docker/ubuntu-noble

produces. Something wth the recent py changes is not perfect:


  => ERROR [28/37] RUN python3 -m pip install --no-cache-dir --upgrade 
"tornado>=6.3.0" "twisted>=24.3.0" "zope.in  0.8s
------
  > [28/37] RUN python3 -m pip install --no-cache-dir --upgrade 
"tornado>=6.3.0" "twisted>=24.3.0" "zope.interface>=6.1":
0.689 error: externally-managed-environment
0.689
0.689 × This environment is externally managed
0.689 ╰─> To install Python packages system-wide, try apt install
0.689     python3-xyz, where xyz is the package you are trying to
0.689     install.
0.689
0.689     If you wish to install a non-Debian-packaged Python package,
0.689     create a virtual environment using python3 -m venv path/to/venv.
0.689     Then use path/to/venv/bin/python and path/to/venv/bin/pip. Make
0.689     sure you have python3-full installed.
0.689
0.689     If you wish to install a non-Debian packaged Python application,
0.689     it may be easiest to use pipx install xyz, which will manage a
0.689     virtual environment for you. Make sure you have pipx installed.
0.689
0.689     See /usr/share/doc/python3.12/README.venv for more information.
0.689
0.689 note: If you believe this is a mistake, please contact your Python 
installation or OS distribution provider. You can override this, at the 
risk of breaking your Python installation or OS, by passing 
--break-system-packages.
0.689 hint: See PEP 668 for the detailed specification.
------

  15 warnings found (use docker --debug to expand):
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 148)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 309)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 78)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 85)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 110)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 177)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 269)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 156)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 157)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 284)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 97)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 126)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 155)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 109)
  - LegacyKeyValueFormat: "ENV key=value" should be used instead of 
legacy "ENV key value" format (line 130)
Dockerfile:258
--------------------
  256 |           python3-wheel
  257 |
  258 | >>> RUN python3 -m pip install --no-cache-dir --upgrade 
"tornado>=6.3.0" "twisted>=24.3.0" "zope.interface>=6.1"
  259 |
  260 |     RUN apt-get install -y --no-install-recommends \
--------------------
ERROR: failed to build: failed to solve: process "/bin/sh -c python3 -m 
pip install --no-cache-dir --upgrade \"tornado>=6.3.0\" 
\"twisted>=24.3.0\" \"zope.interface>=6.1\"" did not complete 
successfully: exit code: 1

The env key stuff can be ignored for now. But the error is a blocker.

Thanks for any help,
JensG



Am 07.04.2026 um 14:20 schrieb Carel Combrink:
> Hi Jens,
>
> I see my last PR was merged, thanks for that.
>
> Any progress on the release?
>
> Please shout if there is something that I can do to help if there is 
> something that needs a hand.
>
> Regards,
> Carel

Re: Next Release

Posted by Carel Combrink <ca...@gmail.com>.
Hi Jens,

I see my last PR was merged, thanks for that.

Any progress on the release?

Please shout if there is something that I can do to help if there is
something that needs a hand.

Regards,
Carel

Re: Next Release

Posted by Dmytro Shteflyuk <kp...@kpumuk.info>.
 Hello Jens,

I am working on a two large changes for ruby (TConfiguration and Zeitwerk),
which would make sense to postpone until the next release. Besides a small
bug in SSL connection handling, there are no concerns for Ruby library, and
we have already deployed it to production from a master snapshot — no
issues observed.

I would like to switch Ruby to Trusted Publisher pipeline for the next
release
as well, as I don't want to block the 0.23.0 release.
https://guides.rubygems.org/trusted-publishing/

Good luck

-- 
Dmytro Shteflyuk

On March 27, 2026 at 7:02:36 AM, Jens Geyer (jensgeyer@hotmail.com) wrote:

Hi,

I would reactivate that thread again. Depending on the time slots
available, I will create a release branch in the next days or so., after
my two pending PRs got merged. If there are any blockers to a release
0.23.0, this is right the time to raise your hand.

Have fun,
JensG


Am 16.11.2025 um 12:48 schrieb Jens Geyer:
> Hi all,
>
> it is that time of the year again and so I would start putting a
> 0.23.0 RC together somewhere in the next two weeks.
>
> If there are any objections, e.g. because we want to fix/integrate
> something first that should make it into that release, just let me/us
> know. The date is not set in concrete, it would simply be nice to know
> early.
>
> Have fun,
> JensG
>
>
>

Re: Next Release

Posted by Jens Geyer <je...@hotmail.com>.
Hi,

I would reactivate that thread again. Depending on the time slots 
available, I will create a release branch in the next days or so., after 
my two pending PRs got merged. If there are any blockers to a release 
0.23.0, this is right the time to raise your hand.

Have fun,
JensG


Am 16.11.2025 um 12:48 schrieb Jens Geyer:
> Hi all,
>
> it is that time of the year again and so I would start putting a 
> 0.23.0 RC together somewhere in the next two weeks.
>
> If there are any objections, e.g. because we want to fix/integrate 
> something first that should make it into that release, just let me/us 
> know. The date is not set in concrete, it would simply be nice to know 
> early.
>
> Have fun,
> JensG
>
>
>