Thanks! I liked Christian's initial efforts in #96001; thanks for picking this up.

Instead of the grep exercises; what do you think of the printf approach taken in the CMake-for-CPython project: https://github.com/python-cmake-buildsystem/python-cmake-buildsystem/blob/master/cmake/platform.c

Instead of the grep exercises; what do you think of the printf approach taken in the CMake-for-CPython project: https://github.com/python-cmake-buildsystem/python-cmake-buildsystem/blob/master/cmake/platform.c

That would require fully compiling the program (the current approach only requires the preprocessing step) and running it to get the output, which would not be feasible when cross-compiling.

My preference would be something similar to config.guess, where we output the triplet with a prefix like PLATFORM_TRIPLET= then grep for the prefix specifically (but without evaling the output). Would this be acceptable?

That would require fully compiling the program (the current approach only requires the preprocessing step) and running it to get the output, which would not be feasible when cross-compiling.

That's a good point.

My preference would be something similar to config.guess, where we output the triplet with a prefix like PLATFORM_TRIPLET= then grep for the prefix specifically (but without evaling the output). Would this be acceptable?

Yeah, I would prefer something like that.

Yeah, I would prefer something like that.

Updated 👍

🤖 New build scheduled with the buildbot fleet by @erlend-aasland for commit 7fb142c 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

Not sure if the buildbot errors are related to this change?

Not sure if the buildbot errors are related to this change?

Yeah, unfortunately we'll have to examine each bot. I'll see if I can find time later. Feel free to take a look yourself :)

I have looked at every error log - while the errors do not look related to this change to me, I'm not familiar enough with the tests being run to say for certain.

Is there anything else I can do to help resolve the buildbot errors (perhaps update/rebase the branch)?

I did a fairly quick inspection of the buildbot run. Feel free to go through them as well (or even better a different subset).

As you see, there are some changes.

Also, I noticed this configure warning in the Gentoo Non-Debug buildbot (despite it getting the i386-linux-gnu triplet):

configure: WARNING: i686-pc-linux-gnu/gcc is not supported

• PPC64LE * (was powerpc64le-linux-gnu, now powerpc-linux-gnu)
  This should be fixed
• PPC64 Fedora PR (was powerpc-linux-gnu, now powerpc64-linux-gnu) (build)
  I think the results are switched (was powerpc64-linux-gnu, now powerpc-linux-gnu) - the change is the same issue as with PPC64LE (and should also be fixed)
• x86 Gentoo Installed with X PR (was i686-pc-linux-gnu, now i386-linux-gnu) (build)
  I checked several previous builds and they all found i386-linux-gnu as the platform triplet (with i686-pc-linux-gnu as the build/host system type)

🤖 New build scheduled with the buildbot fleet by @AA-Turner for commit cbaf416 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

@jefferyto I've kicked off another buildbot run for your latest changes.

A

@AA-Turner thanks!

Thanks, Adam! @jefferyto, can you do an audit of the buildbot run when it completes? I won't be able to return to this until probably late next week.

Here are all the results:

I found no regressions, though I suppose this will need to be independently verified.

Wow, good job! Thank you so much. I'll try to find time to verify it the coming week.

@erlend-aasland ping 🙏

@corona10, do you want to have a look?

@tiran, if you are around; feel free to have a look :)

Okay I will take a look but please wait until this weekend ;)

Okay I will take a look but please wait until this weekend ;)

Sure, there is no hurry :)

cc. @indygreg

Thanks @jefferyto, Christian, and Dong-hee!

Thanks @erlend-aasland for your help as well!