[release] Python 3 build problems - Boost - lists.stage.boost.cppalliance.org

newer
[ValueRef] Interest in a new...

[release] Python 3 build problems

older
[regex] Support for Perl's (*SKIP)

Daniel James

17 Apr 2015 17 Apr '15

11:30 a.m.

Hi, Is there anyone who can write a brief description of the build problem with python 3 for the release notes? We probably should say something. thanks, Daniel

Reply

Sign in to reply online Use email software

Show replies by date

Stefan Seefeld

17 Apr 17 Apr

11:58 a.m.

On 17/04/15 07:30 AM, Daniel James wrote:

Hi,

Is there anyone who can write a brief description of the build problem with python 3 for the release notes? We probably should say something.

I checked in a patch a few weeks ago that should have fixed the issue that was raised recently. Is this not included in the release ? If so, why not ? Or are we talking about a different issue here ? Thanks, Stefan -- ...ich hab' noch einen Koffer in Berlin...

Reply

Sign in to reply online Use email software

Daniel James

12:26 p.m.

On 17 April 2015 at 12:58, Stefan Seefeld wrote:

On 17/04/15 07:30 AM, Daniel James wrote:

...
Hi,

Is there anyone who can write a brief description of the build problem with python 3 for the release notes? We probably should say something.

I checked in a patch a few weeks ago that should have fixed the issue that was raised recently. Is this not included in the release ? If so, why not ? Or are we talking about a different issue here ?

The release is based on RC3, and there was some sort of build problem with that. That's pretty much all I know, I haven't been following this issue.

Reply

Sign in to reply online Use email software

Tom Kent

2:30 p.m.

On Fri, Apr 17, 2015 at 6:30 AM, Daniel James wrote:

Hi,

Is there anyone who can write a brief description of the build problem with python 3 for the release notes? We probably should say something.

I think a summary might be: Boost::Python will fail to build if it is complied agains a version of python that is one of: 3.0.X, 3.1.X, 3.2.X, 3.3.X. Versions 2.X and 3.4+ are not affected. Tom

Reply

Sign in to reply online Use email software

Daniel James

5:57 p.m.

On 17 April 2015 at 15:30, Tom Kent wrote:

Boost::Python will fail to build if it is complied agains a version of python that is one of: 3.0.X, 3.1.X, 3.2.X, 3.3.X. Versions 2.X and 3.4+ are not affected.

Thanks. Btw. I can also add a patch to the release notes if needed.

Reply

Sign in to reply online Use email software

Stefan Seefeld

6:02 p.m.

On 17/04/15 01:57 PM, Daniel James wrote:

On 17 April 2015 at 15:30, Tom Kent wrote:

...
Boost::Python will fail to build if it is complied agains a version of python that is one of: 3.0.X, 3.1.X, 3.2.X, 3.3.X. Versions 2.X and 3.4+ are not affected. Thanks. Btw. I can also add a patch to the release notes if needed.

The patch to fix the issue for Python3 < 3.4 is https://github.com/boostorg/python/commit/3e405b6fd5db5615bbef241763de070118... FWIW, Stefan PS: It's a pitty that none of the test builders indicate what Python version they are running with. That would have been valuable information for Boost.Python tests. -- ...ich hab' noch einen Koffer in Berlin...

Reply

Sign in to reply online Use email software

Rene Rivera

6:22 p.m.

On Apr 17, 2015 1:03 PM, "Stefan Seefeld" wrote:

On 17/04/15 01:57 PM, Daniel James wrote:

...
On 17 April 2015 at 15:30, Tom Kent wrote:

...
Boost::Python will fail to build if it is complied agains a version of python that is one of: 3.0.X, 3.1.X, 3.2.X, 3.3.X. Versions 2.X and

3.4+

...
...
are not affected. Thanks. Btw. I can also add a patch to the release notes if needed.

The patch to fix the issue for Python3 < 3.4 is

https://github.com/boostorg/python/commit/3e405b6fd5db5615bbef241763de070118...

FWIW, Stefan

PS: It's a pitty that none of the test builders indicate what Python version they are running with. That would have been valuable information for Boost.Python tests.

If it's important you should add an informational test that shows the Python version being tested.

Reply

Sign in to reply online Use email software

Stefan Seefeld

6:26 p.m.

On 17/04/15 02:22 PM, Rene Rivera wrote:

On Apr 17, 2015 1:03 PM, "Stefan Seefeld" wrote:

...
PS: It's a pitty that none of the test builders indicate what Python version they are running with. That would have been valuable information for Boost.Python tests. If it's important you should add an informational test that shows the Python version being tested.

While that might work, it seems a bit hackish to me. It seems the Python version (as well as a number of other config parameters) are valuable metadata associated with a given build that would be useful to display, alongside the compiler name / version in use. Stefan -- ...ich hab' noch einen Koffer in Berlin...

Reply

Sign in to reply online Use email software

Rene Rivera

6:51 p.m.

On Apr 17, 2015 1:27 PM, "Stefan Seefeld" wrote:

On 17/04/15 02:22 PM, Rene Rivera wrote:

...
On Apr 17, 2015 1:03 PM, "Stefan Seefeld" wrote:

...
PS: It's a pitty that none of the test builders indicate what Python version they are running with. That would have been valuable

information

...
...
for Boost.Python tests. If it's important you should add an informational test that shows the Python version being tested.

While that might work, it seems a bit hackish to me. It seems the Python version (as well as a number of other config parameters) are valuable metadata associated with a given build that would be useful to display, alongside the compiler name / version in use.

It's actually the less hackish way. As it handles the eventuality that the version of Python that the tools are using will be different than the one being tested for BPL. And would also account for multiple versions being tested in the same b2 run.

Reply

Sign in to reply online Use email software

Tom Kent

20 Apr 20 Apr

1:53 p.m.

On Fri, Apr 17, 2015 at 1:51 PM, Rene Rivera wrote:

On Apr 17, 2015 1:27 PM, "Stefan Seefeld" wrote:

...
On 17/04/15 02:22 PM, Rene Rivera wrote:

...
On Apr 17, 2015 1:03 PM, "Stefan Seefeld" wrote:

...
PS: It's a pitty that none of the test builders indicate what Python version they are running with. That would have been valuable

information

...
...
...
for Boost.Python tests. If it's important you should add an informational test that shows the Python version being tested.

While that might work, it seems a bit hackish to me. It seems the Python version (as well as a number of other config parameters) are valuable metadata associated with a given build that would be useful to display, alongside the compiler name / version in use.

It's actually the less hackish way. As it handles the eventuality that the version of Python that the tools are using will be different than the one being tested for BPL. And would also account for multiple versions being tested in the same b2 run.

I'd agree with this, if we can make it part of the test. However, it might be nice to include some more information automatically in the comment html file, just like we do with the command that is executed. For debugging purposes it might be nice to see: contents of the user-config.jam file, version of python that run.py is executed with, actual version of the toolsets being used (can we get this from bjam?). I know when I was getting my runners first setup, I had tons of problems with my user-config.jam file (and still do from time to time), and if others could have seen that in my test setup and mentioned it to me, things could have been running more smoothly much sooner. Tom

Reply

Sign in to reply online Use email software

Rene Rivera

2:12 p.m.

On Mon, Apr 20, 2015 at 8:53 AM, Tom Kent wrote:

On Fri, Apr 17, 2015 at 1:51 PM, Rene Rivera wrote:

...
On Apr 17, 2015 1:27 PM, "Stefan Seefeld" wrote:

...
On 17/04/15 02:22 PM, Rene Rivera wrote:

...
On Apr 17, 2015 1:03 PM, "Stefan Seefeld"

wrote:

...
...
...
PS: It's a pitty that none of the test builders indicate what Python version they are running with. That would have been valuable

information

...
...
for Boost.Python tests. If it's important you should add an informational test that shows the Python version being tested.

While that might work, it seems a bit hackish to me. It seems the Python version (as well as a number of other config parameters) are valuable metadata associated with a given build that would be useful to display, alongside the compiler name / version in use.

It's actually the less hackish way. As it handles the eventuality that the version of Python that the tools are using will be different than the one being tested for BPL. And would also account for multiple versions being tested in the same b2 run.

I'd agree with this, if we can make it part of the test.

However, it might be nice to include some more information automatically in the comment html file, just like we do with the command that is executed.

For debugging purposes it might be nice to see: contents of the user-config.jam file, version of python that run.py is executed with, actual version of the toolsets being used (can we get this from bjam?).

I would avoid including user-config.jam or system-config.jam in the uploaded information. They are "private" files and may contain confidential information. And hence we would be liable for posting such information on a publicly accessible location. We can definitely include version information though. And maybe scrape the --debug-configuration information from BB and put it in the info (it should have the toolset information). -- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail

Reply

Sign in to reply online Use email software

Rene Rivera

2:15 p.m.

On Mon, Apr 20, 2015 at 9:12 AM, Rene Rivera wrote:

On Mon, Apr 20, 2015 at 8:53 AM, Tom Kent wrote:

...
On Fri, Apr 17, 2015 at 1:51 PM, Rene Rivera wrote:

...
On Apr 17, 2015 1:27 PM, "Stefan Seefeld" wrote:

...
On 17/04/15 02:22 PM, Rene Rivera wrote:

...
On Apr 17, 2015 1:03 PM, "Stefan Seefeld"

...
...
...
PS: It's a pitty that none of the test builders indicate what

Python

...
...
version they are running with. That would have been valuable information for Boost.Python tests. If it's important you should add an informational test that shows

wrote: the

...
...
Python version being tested.

While that might work, it seems a bit hackish to me. It seems the Python version (as well as a number of other config parameters) are valuable metadata associated with a given build that would be useful to display, alongside the compiler name / version in use.

It's actually the less hackish way. As it handles the eventuality that the version of Python that the tools are using will be different than the one being tested for BPL. And would also account for multiple versions being tested in the same b2 run.

I'd agree with this, if we can make it part of the test.

However, it might be nice to include some more information automatically in the comment html file, just like we do with the command that is executed.

For debugging purposes it might be nice to see: contents of the user-config.jam file, version of python that run.py is executed with, actual version of the toolsets being used (can we get this from bjam?).

I would avoid including user-config.jam or system-config.jam in the uploaded information. They are "private" files and may contain confidential information. And hence we would be liable for posting such information on a publicly accessible location.

We can definitely include version information though. And maybe scrape the --debug-configuration information from BB and put it in the info (it should have the toolset information).

PS. If you have specific feature to add to the regression testing please file a task in the github boostorg/regression repo < https://github.com/boostorg/regression/issues>. So that I have a ready list of stuff to do when I have free time.

-- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail

Reply

Sign in to reply online Use email software

Tom Kent

17 Apr 17 Apr

6:49 p.m.

On Fri, Apr 17, 2015 at 1:02 PM, Stefan Seefeld wrote:

PS: It's a pitty that none of the test builders indicate what Python version they are running with. That would have been valuable information for Boost.Python tests.

I've just started to add it to mine: http://www.boost.org/development/tests/develop/developer/summary.html Should be rolling out to the teeks99* (at least the windows ones) over the next couple weeks. Tom

Reply

Sign in to reply online Use email software

3509

Age (days ago)

3512

Last active (days ago)

Download

12 comments

4 participants

tags

participants (4)

Daniel James
Rene Rivera
Stefan Seefeld
Tom Kent