Poll: Intel 32bit (i686) builds (linux and/or windows)

classic Classic list List threaded Threaded
7 messages Options
Reply | Threaded
Open this post in threaded view
|

Poll: Intel 32bit (i686) builds (linux and/or windows)

Sven Gothel
Administrator
Poll: Intel 32bit (i686) builds (linux and/or windows)

See
http://forum.jogamp.org/JogAmp-RC-Build-2-4-0-rc-20200102-tp4040226p4040229.html

For now these platforms have been disabled,
while still build-able of course (source code support).

MacOS has no more support for i686.

Who would like to have linux-x86 and windows-x86 being
still build and provided by (RC) releases?

Also please state the reason, i.e. describing
your desired target platform.

Thank you.

~Sven
Reply | Threaded
Open this post in threaded view
|

Re: Poll: Intel 32bit (i686) builds (linux and/or windows)

Manu
Thanks to the update system programmed in Sweet Home 3D, I can check whether its users run the 32 or 64 bit version of the program. And just for yesterday, I can tell you that almost 10000 users were using the 32 bit version of that program.
Hope that this will be a good enough reason to maintain the 32 bit version of JOGL (even if some users of the 32 bit version might be able to able to run the 64 bit version too if their system permits it).

Many thanks for your efforts and happy new year 2020!
Emmanuel Puybaret
Reply | Threaded
Open this post in threaded view
|

Re: Poll: Intel 32bit (i686) builds (linux and/or windows)

Sven Gothel
Administrator
Thank you Emmanuel for replying.

I assume your 32bit users were mostly using windows-x86?

Thinking more about linux-x86, it might be still an important
platform at least for embedded devices hmm.

I will go for a compromise here:

- Re-enabling linux-x86 and windows-x86 builds and aggregation,
  but skipping the time and energy consuming unit tests.

- The unit tests on 32bit went less reliable on JOGL
  running on actual 64bit machines and OS due to GL driver glitches.

- I recommend to run 64bit on 64bit machines

This shall be a good compromise and from time to time
I could enable unit testing (a flag in jenkins) on these machines.

All 32bit users should earmark this situation and
maybe also perform the unit tests on their target platform etc.

I would like to read more use cases here, maybe we get more
communication when everybody is back :)

Cheers, Sven

On 1/3/20 12:30 AM, Manu [via jogamp] wrote:

> Thanks to the update system programmed in Sweet Home 3D, I can check whether
> its users run the 32 or 64 bit version of the program. And just for yesterday,
> I can tell you that almost 10000 users were using the 32 bit version of that
> program.
> Hope that this will be a good enough reason to maintain the 32 bit version of
> JOGL (even if some users of the 32 bit version might be able to able to run
> the 64 bit version too if their system permits it).
>
> Many thanks for your efforts and happy new year 2020!
> Emmanuel Puybaret
>
Reply | Threaded
Open this post in threaded view
|

Re: Poll: Intel 32bit (i686) builds (linux and/or windows)

Sven Gothel
Administrator
Debian Popularity Contest <https://popcon.debian.org/>'s
"Statistics per Debian architectures - submissions per arch" shows

- amd64 174836
- i386   20318
- armhf    585
- armel    407
- powerpc  133
- arm64     98

This surely justifies a linux-x86 (i386) build.

Thanks to Xerxes, who put my nose to this dataset.

~Sven

On 1/4/20 1:02 AM, Sven Gothel [via jogamp] wrote:

> Thank you Emmanuel for replying.
>
> I assume your 32bit users were mostly using windows-x86?
>
> Thinking more about linux-x86, it might be still an important
> platform at least for embedded devices hmm.
>
> I will go for a compromise here:
>
> - Re-enabling linux-x86 and windows-x86 builds and aggregation,
>   but skipping the time and energy consuming unit tests.
>
> - The unit tests on 32bit went less reliable on JOGL
>   running on actual 64bit machines and OS due to GL driver glitches.
>
> - I recommend to run 64bit on 64bit machines
>
> This shall be a good compromise and from time to time
> I could enable unit testing (a flag in jenkins) on these machines.
>
> All 32bit users should earmark this situation and
> maybe also perform the unit tests on their target platform etc.
>
> I would like to read more use cases here, maybe we get more
> communication when everybody is back :)
>
> Cheers, Sven
>
> On 1/3/20 12:30 AM, Manu [via jogamp] wrote:
>
>> Thanks to the update system programmed in Sweet Home 3D, I can check whether
>> its users run the 32 or 64 bit version of the program. And just for yesterday,
>> I can tell you that almost 10000 users were using the 32 bit version of that
>> program.
>> Hope that this will be a good enough reason to maintain the 32 bit version of
>> JOGL (even if some users of the 32 bit version might be able to able to run
>> the 64 bit version too if their system permits it).
>>
>> Many thanks for your efforts and happy new year 2020!
>> Emmanuel Puybaret
>>
Reply | Threaded
Open this post in threaded view
|

Re: Poll: Intel 32bit (i686) builds (linux and/or windows)

Sven Gothel
Administrator
jenkins nodes have linux-i586 and windows-i586
back, building the binaries.

~Sven
Reply | Threaded
Open this post in threaded view
|

Re: Poll: Intel 32bit (i686) builds (linux and/or windows)

sailsman63
In reply to this post by Sven Gothel
I've been trying to get a response from our user community, just to gauge things. Not much so far. The one thing that I have seen happen in the past is that some windows users end up with 32-bit jre even though they have 64 bit windows.

(Usually due to having a 32bit build of their browser of choice. Java.com probes your browser for a recommended default download)
Reply | Threaded
Open this post in threaded view
|

Re: Poll: Intel 32bit (i686) builds (linux and/or windows)

gouessej
Administrator
sailsman63 wrote
I've been trying to get a response from our user community, just to gauge things. Not much so far. The one thing that I have seen happen in the past is that some windows users end up with 32-bit jre even though they have 64 bit windows.

(Usually due to having a 32bit build of their browser of choice. Java.com probes your browser for a recommended default download)
You shouldn't rely on the JRE installed on the system anyway.

Personally, I'm in favor of dropping 32-bit support if and only if we ensure that we don't require any fancy CPU instruction on 64-bit. Hardware supporting 32-bit CPU is becoming very difficult to repair, even my oldest computers with 64-bit CPU are difficult to repair because of the lack of spare parts. Anyway, maybe it's too early to do it.
Julien Gouesse | Personal blog | Website