leon-anavi has quit [Remote host closed the connection]
leon-anavi has joined #yocto
rfuentess has quit [Read error: Connection reset by peer]
rfuentess has joined #yocto
belsirk has joined #yocto
rfuentess has quit [Ping timeout: 252 seconds]
rfuentess has joined #yocto
belsirk has quit [Ping timeout: 260 seconds]
tlwoerner has quit [Ping timeout: 260 seconds]
tlwoerner has joined #yocto
tlwoerner has quit [Ping timeout: 244 seconds]
tlwoerner has joined #yocto
rber|res has joined #yocto
ablu has quit [Ping timeout: 252 seconds]
ablu has joined #yocto
tgamblin has joined #yocto
<rburton>
khem: will you be expediting the wxwidgets fix for meta-oe from jama?
florian has quit [Ping timeout: 260 seconds]
thomas_34 has quit [Quit: Client closed]
florian has joined #yocto
leonanavi has joined #yocto
leon-anavi has quit [Ping timeout: 252 seconds]
ptsneves has quit [Ping timeout: 276 seconds]
sakoman has quit [Ping timeout: 276 seconds]
rfuentess has quit [Quit: freaking vpns]
<LetoThe2nd>
I'm trying to experiment with yocto-check-layer, and it immediately fails with INFO: Skipping meta-sulka-distro due to missing dependencies., albeit being run from a functional build setup which has the layer dependencies enabled. where am I possibly going wrong?
kbo has joined #yocto
<RP>
LetoThe2nd: try including the path to the dependent layers even if they are configured
<rburton>
i prefer using yocto-check-layer-wrapper as it uses a fresh build tree, even if it makes calling the command even harder
<LetoThe2nd>
rburton: let me look at that, this really goes sideways quick. one of the dependencies is meta-security, which in turn contains meta-parsec, which wants meta-clang but that isn't around - because I nowhere need or use meta-parsec...
<rburton>
there's actually a bug about how its a terrible tool to drive
ptsneves has joined #yocto
<LetoThe2nd>
definitely not getting anywhere :-(
rfuentess has joined #yocto
tgamblin has quit [Quit: WeeChat 4.6.3]
paulg has joined #yocto
ptsneves has quit [Ping timeout: 248 seconds]
warthog9 has quit [Ping timeout: 272 seconds]
warthog9 has joined #yocto
cyxae has joined #yocto
tgamblin has joined #yocto
sakoman has joined #yocto
Daanct12 has quit [Quit: WeeChat 4.6.3]
Xagen has joined #yocto
rob_w has quit [Remote host closed the connection]
jmd has joined #yocto
<fullstop>
I have two machines with similar images. One image gets the python3 ensurepip module but the other does not, yet they have the same python packages installed. Any idea why one is different, or how can I track this down?
<fullstop>
The only difference that I can spot is that I have appended python3-dev to the TOOLCHAIN_TARGET_TASK
<RP>
LetoThe2nd: the interface to that tool does need work
<rburton>
RP: can i do something magic in a target recipe's sysroot preprocess function to put a _native_ sysroot path in a generated file?
<rburton>
(clang's target llvmconfig.cmake has "directory where binaries are" which should be the native sysroot not the target sysroot)
<fullstop>
odd, the graphvis dot file does not show python3-ensurepip for either image, yet it is definitely in one of them.
druppy has quit [Read error: Connection reset by peer]
druppy has joined #yocto
<rburton>
fullstop: have you compared the image manifests to verify the package lists are actually the same?
kbo has quit [Quit: Client closed]
<fullstop>
I can check the manifest for the original image, but I just added python3-ensurepip to the second as a test.
<rburton>
like, if you want ensurepip in an image, adding it seems like the right thing to do :)
<fullstop>
rburton: it is pulled in as a dependency somehow, maybe with python3-pip ?
<fullstop>
It is in the original image's manifest. I have never explicitly included it.
<rburton>
ironically, pip doesn't depend on it
<rburton>
fullstop: i'm guessing something else pulled it in. maybe one of your images actually pulls in python3 the metapackage
<fullstop>
maybe python3-venv ?
<fullstop>
To be clear -- I have python3 and python3-venv, but I am unable to create a venv because the ensurepip module is missing.
<rburton>
python3-modules should be depending on python3-ensurepip (looking at master)
<rburton>
python3-modules also provides: python3
<fullstop>
I am not explicitly pulling in python3-modules
<rburton>
you said you installed python3, so that would mean python3-modules
<fullstop>
The python packages included are almost the same, but one image has python3-opencv and python3-pillow. That image has ensurepip
<fullstop>
I don't explicitly include python3, it seems. I guess I was relying on the modules to pull it in.
<rburton>
unless you explicitly install python3 then you get the subset of the library that your dependencies ask for
<fullstop>
Okay, I will add python3 and see if ensurepip comes in
<rburton>
if pip actually needs python3-ensurepip then send a patch to add it to the pip recipe
<rburton>
RP: libclc split sent - means no need to build target clc in situations where its not needed and pulls another recipe out of the dependency chain for clang itself
<fullstop>
"Unless the --without-pip option is given, ensurepip will be invoked to bootstrap pip into the virtual environment."
druppy has quit [Read error: Connection reset by peer]
<rburton>
i'd err on the side of caution as its small and just pull it in anyway as a depends
<fullstop>
I understand. I was just surprised that it was missing, but I should just be more explicit if I'm actively using it
<rburton>
well both. its missing because nobody who noticed has sent a patch to fix it
<rburton>
you can be the change you want in the world!
<fullstop>
This is probably why I had to add python3-dev to the toolchain task as well.
<fullstop>
rburton: I will try to create a patch. The lack of git integration with gmail is something of a roadblock when submitting patches, but maybe I was doing things wrong.
astlep5504018066 has quit [Quit: Ping timeout (120 seconds)]
<rburton>
you can use b4 to avoid email entirely, or its absolutely possible to smtp to gmail
frieder has quit [Remote host closed the connection]
<fullstop>
roger
astlep5504018066 has joined #yocto
<rburton>
the trick with gmail is creating an app password, so you don't need to 2fa (which git can't do)
<RP>
rburton: wouldn't you just hack the right thing in during a do_configure prefunc ?
<fullstop>
I ran into that problem and didn't want to disable my yubikey stuff
<fullstop>
I'll look at app passwords
<rburton>
fullstop: yeah app passwords are the solution there
<rburton>
RP: would have to be done for everything that uses it
<fullstop>
I can also confirm that including python3 gets ensurepip installed
<RP>
rburton: annoying but the target sysroot is not really designed to have native paths :(
<RP>
rburton: can it take relative paths?
<rburton>
i _guess_ so
<RP>
why would 25% of our cache load time for a valid cache be in _int_to_enum in signal.py
* RP
suspects we're doing something silly
<RP>
ouch, it comes from pthread_sigmask :(
<fullstop>
rburton: I am a little more confused now, since python has --without-ensurepip as part of the EXTRA_OECONF arguments.
<rburton>
RP: thread thrashing or something?
<RP>
rburton: our code does some signal masking for lock protection but the code in 3.13 python is not really performance optimised
<RP>
return set(_int_to_enum(x, Signals) for x in sigs_set)
<rburton>
erm that seems suboptimal
<RP>
that equates to over 800,000 calls to _init_to_enum taking 1.25s
<rburton>
monkeypatch that to something better?
<RP>
rburton: I will seek advice from someone who understands python better than I to!
<rburton>
if you can import _signal directly then you can avoid the set nonsense and call the lower function
<rburton>
PSA call for testers. does anyone actually use the GL passthrough/qemu stuff? ie run a qemux86 image inside qemu and it does GL magic rendering using the host GPU?