<Guest86>
I want to use a change/patch made in oe-core as a template in a meta-layer, what can I do to give credit to the person who made it in oe-core? I will not make the exactly the same change but I will use the same commit message and subject for example.
<rburton>
just reference them in the commit message, git has no markup for 'originally by but altered'
<Guest86>
thanks
Guest86 has quit [Quit: Client closed]
Kbo has joined #yocto
<CrazyGecko>
we're updating to scarthgap and we put the licences together while building the image and put it into the rootfs. with the commit 75ac1f529a22e56988bda53b58d2cc34fb7baea4 in poky, the path of the licences change and is now put under the `SSTATE_PKGARCH` / `PACKAGE_ARCH` folder.
<CrazyGecko>
Currently we rely on the packagedata to get the used license, and did put together the license path with "os.path.join(d.getVar('LICENSE_DIRECTORY'),pkg_data["PN"])". Now we should add in between the PACKAGE_ARCH. Would it make sense to store the PACKAGE_ARCH also in the packagedata or does someone know another way to get now the path to the
<CrazyGecko>
licence file? Our workaround would be to search in the LICENSE_DIRECTORY
tgamblin has quit [Quit: WeeChat 3.8]
tgamblin has joined #yocto
GNUmoon has quit [Ping timeout: 244 seconds]
GNUmoon has joined #yocto
druppy has quit [Read error: Connection reset by peer]
<rburton>
can't you just tell everyone to stop using it ;)
Pinta has joined #yocto
druppy has quit [Read error: Connection reset by peer]
<vmeson>
rburton: system moslty idle. Added meta-openembedded/meta-* layers, My test improved: 192 = 30s, 128= 30s, 96= 30s, 64 = 31s, 48= 34s, 32 = 36s, 16 = 47s, 8 = 70s, 4= too long! Running: -- rm -rf cache/ sstate-cache/ tmp/; time bitbake -p
Guest70 has joined #yocto
<Guest70>
Some of the recipes using PV .= "+1.0+git" something like that. Is that necessary?
<Guest70>
It seems unnecessary. There is already a version in recipe name.
<Guest70>
Am I wrong?
florian_kc is now known as florian
<rburton>
Guest70: in core, walnascar has mostly dropped those apart from a few recipes where the +git means its literally a snapshot from git and not a release
<rburton>
historically that was how a git recipe was versioned but yes, if the sha is a tag then its redundant information, which is why its been removed
<Guest70>
rburton thanks
Guest70 has quit [Quit: Client closed]
cyxae has quit [Quit: cyxae]
<tgamblin>
rburton: have you ever considered putting python-unittest-automake-output on PyPI?
* tgamblin
just ran into a scenario where it would've been useful to install with pip
<rburton>
tgamblin: yeah it was on my todo list
<rburton>
it needs a better name first ;)
<RP>
rburton: hmm. Most of it is fine, I'm just not sure about the 64 limit by default for parsing
<RP>
rburton: I'm curious of the bottleneck on your parsing with 128 too
<rburton>
remind me tomorrow, i'll add more layers and script getting times
<tgamblin>
rburton: pyaso for 'python automake style output' maybe?
goliath has quit [Quit: SIGSEGV]
<RP>
rburton: ok, will do
jmd has quit [Remote host closed the connection]
Guest12 has joined #yocto
<Guest12>
I'm trying to fetch a project in SRC_URI but project trying to init submodule in do_configure step. How can I deal with this?
zeddii has quit [Ping timeout: 260 seconds]
<RP>
Guest12: change it to use the gitsm fetcher?
<RP>
Guest12: prefetch the submodules in SRC_URI so it doesn't have to do that?
<RP>
sometimes you can disable the submodules too
<Guest12>
RP Tried with gitsm and still error occurs. I think it's related to the project.
<Guest12>
RP I think it can't fetch because of the SSH url in .gitmodules. Since the submodule in .gitmodules is defined as git@github.com:... (SSH), it gives error again when trying to fetch with HTTPS.
zeddii has joined #yocto
<RP>
Guest12: we disable network access in do_configure which probably doesn't help too
<Guest12>
What should I do in this situation? Do you have any suggestion?
jmiehe has joined #yocto
<RP>
Well, I already suggested 3 different options
<RP>
you could enable network access for do_configure but understand why its a really really bad idea
dmoseley has joined #yocto
<RP>
JPEW: I've put a tweaked version of the parser patch in master-next, basically dropping one of the safety nets for jobs in favour of the other ones we have
<JPEW>
I think I had a bit of an epipinany in a better way to do it also; hang on and I'll push what I have
<RP>
JPEW: I just removed the "havejobs" variable entirely and relied on the self.exit and self.quit functioning being enough
<JPEW>
TL; DR instead of using a queue to feed the parser processes, use a shared counter
<RP>
JPEW: That is a neat idea but you do remove most of the sanity protections that have been added in there for various reasons over time :/
<RP>
JPEW: the code does look a bit paranoid about shutdown in particular but there is good reason for it :(
Guest12 has quit [Quit: Client closed]
<JPEW>
Ya, you can keep that if you want; it should still work
<RP>
JPEW: I'll see how the current testing goes and if that doesn't work I'll see how this looks
<RP>
I just remember how painful some of this was to debug and work out in the first place :/
ablu has quit [Remote host closed the connection]
ablu has joined #yocto
<JPEW>
RP: I split it into 2 patches; one does the shared counter, the other has just the process cleanup rework
<RP>
JPEW: it still blocks on the push to the results queue though :/
<RP>
That can be changed too but we did find that was a problem :/
<JPEW>
Ya... If my code archaeology is correct that was 2013 (or possibly earlier); the code didn't even bother to try and drain the result_queue to unstick the workers back then
<RP>
JPEW: right, I'm probably just nervous about breaking this again