**** BEGIN LOGGING AT Mon Oct 21 02:59:58 2019 **** BEGIN LOGGING AT Mon Oct 21 06:44:26 2019 Oct 21 07:45:26 good morning Oct 21 16:04:07 khem: ping Oct 21 18:00:10 does anyone think it would be appropriate to modify the distutils class to pass along parallel make flags for python recipes that build native code? Oct 21 18:00:50 i'm interested in contributing this change, but it's my first time. Oct 21 18:11:32 Do we ahve any auto python depends determination story? Oct 21 18:31:33 Crofton|work: not yet. been meaning to look at rpmdeps to see if that can do it Oct 21 18:32:25 Someone at PSEC suggested that would be a huge usability win for us Oct 21 18:33:04 yeah it would Oct 21 18:33:05 ditto, perl Oct 21 18:34:20 mischief: doesn't distutils compile directly? Oct 21 18:34:36 so you'd need to change distutils to spawn multiple workers Oct 21 18:35:31 rburton: it does do that :^) Oct 21 18:35:49 rburton: i have a patch that i can send, but i have to go through some bureacracy first before i can send it. Oct 21 18:36:03 unless you'd like to go do it for me :-D Oct 21 18:36:22 sending upstream to python i presume? Oct 21 18:36:33 no, it's been a feature in distutils for 5 years Oct 21 18:36:43 https://bugs.python.org/issue5309 Oct 21 18:37:55 ah, if theres lots of modules to build Oct 21 18:37:59 sure send the one liner Oct 21 18:38:39 if you've legal pain to beat through convince them to let you just send 'trivial' stuff without oversight, as this is presumably just passing the right -j flag in the right place Oct 21 18:38:49 'look do you want me to bother you every time i send a one line change' Oct 21 18:39:26 (my employer has a policy where trivial patches are fine as long once the project has been approved) Oct 21 18:45:16 yes, that's the boat i'm in Oct 21 18:50:33 mischief, let us know if you need help sorting out policy, we ahv epeople that understand these things Oct 21 18:59:01 Crofton|work: i did have one question, but maybe better for legal Oct 21 19:00:17 We all all amateur lawyers and we know some people doing open source program work for corps we are freindly with Oct 21 19:01:07 does the Developer's Certificate of Origin constitute 'signing an agreement'? Oct 21 19:01:20 hmmm Oct 21 19:01:26 better ask legal Oct 21 19:01:31 :-P Oct 21 19:02:18 https://developercertificate.org/ Oct 21 19:02:22 Developers' Certificate of Origin is documented as a component of contributing.. it says you have to 'sign off' your patches and indicate certain things.. if you can't don't contribute Oct 21 19:02:43 basically, I assume that you have sorted out with your employer that you can submit Oct 21 19:02:51 If you organization won't let you agree to the terms to submit, then yes your legal org has to get involved Oct 21 19:03:19 this is clause 1, basially, I talked to the people who employ me and they are O with me submitting Oct 21 19:04:06 are fine with mme ... Oct 21 19:04:16 yup. exactly Oct 21 19:04:44 or in other words, if they come after us, we through you under this bus :) Oct 21 19:05:17 this is in direct response to SCO suing over various contributions to Linux, that was later shown to have been contributed by SCO.. :P Oct 21 19:06:32 rburton: i suppose i should ask this - is it better to put ${PARALLEL_MAKE} into the distutils_do_compile invocation directly, or use DISTUTILS_BUILD_ARGS? Oct 21 19:32:31 What is the trick to build a kernel-dev packahge you can build modules on target with? Oct 21 19:32:53 I din't see th e/lib/module/version/build directory in image in work dir Oct 21 20:35:36 mischief: i'd just add it directly Oct 21 20:57:06 Crofton|work: recipetool create will do taht using the 'pythondeps' script that's in oe-core. it's not perfect, nothing of the sort is, given it's having to do ast analysis Oct 21 21:02:31 kergoth: i'd like to see it happen like solibs Oct 21 21:06:57 hi folks, FYI the OE TSC meeting is now starting over at #oe-tsc, anyone is welcome to join, minutes will be posted later as well Oct 21 21:08:48 hmm, no beer.. why bother ; ) Oct 21 21:09:14 you are free to crack open your own beer, we won't stop you :D Oct 21 21:09:28 rburton: i think we could use an solibs-based approach that only uses official declared deps in requirements.txt/pyproject.toml or setup_requires, but even the latter can have problems when it changes based on the python version involved Oct 21 21:09:30 i've beer! Oct 21 21:09:57 rburton: the ast-based analysis isn't definitive and requires manual review. when there are try/except conditionals on ImportError, for example, the script picks up the deps on both sides of the conditional Oct 21 21:10:19 right Oct 21 21:10:30 that said, i prototyped it using the current script 4 years ago: https://github.com/kergoth/meta-kergoth-wip/tree/master/meta-package-auto-deps Oct 21 21:10:32 but yeah parsing requirements is actually pretty sensible Oct 21 21:10:39 along with making it more generic for multiple underlying dep mechanisms Oct 21 21:11:47 the python version issue can cause problems even with more standard python tooling. for example, tools like poetry and pipenv that grab the deps and lock them down won't obey per-python-version-deps coming from setup.py, since that's done using actual python code Oct 21 21:11:52 which is why setup.py sucks. pyproject.toml for the win Oct 21 21:11:57 but anyway.. Oct 21 21:23:15 rburton: copy that Oct 21 21:28:29 argh Oct 21 21:28:43 distutils doesn't use icecc to compile X( Oct 21 21:41:41 * mischief is confused why distutils doesn't invoke icecc to build native code when it is clearly enabled **** ENDING LOGGING AT Tue Oct 22 03:00:24 2019