<RP>
paulg: that seems strange unless it just hasn't synced to disk
<paulg>
all the interactions are logged (with DEBUG) as get-outhash. At startup, it indicated anon clients had @dbadmin privs (plus the other defaults)
<paulg>
but I've seen nothing that looks like a write/update db operation triggered by the client...
<RP>
there should be a write after any sstate task completes
<paulg>
so for completeness, I should mention I don't use DL_DIR or SSTATE_DIR ; I just symlink the NFS mounts directly into the build for both (vs. screwing with vars in conf/local.conf). But surely everybody does that.
<RP>
paulg: mostly we set the config but that doesn't matter
bhstalel has quit [Read error: Connection reset by peer]
<paulg>
ok, just checking.
<paulg>
our docs are pretty frustrating with respect to indicating who needs to run hashserv, and how to do it... if you are new to the concept.
<RP>
agreed, I worry that needing to do this isn't clear and how to do it :/
<paulg>
ok - I'll give that a test. 'cause you know, it is only 3:36am and I am stubborn.
<RP>
"Upstream" is a passthrough to something readonly so it matches what you're seeing
<RP>
"auto" means start a local private server
<RP>
you don't want either behaviour
alessioigor has joined #yocto
<paulg>
ok. Our docs really could use a worked example. "Bob has machines A, B, C, and D. A will be exporting downloads and sstate via NFS to B, C and D."
<RP>
paulg: seriously, please send a patch. It is hard to know how to explain this once you know how it works
<paulg>
I get it. I also suck at documenting things I already know how to do. So does everybody.
<luc4>
Hello! I'm trying to debug a layer where it seems some files are missing in the boot partition. What is the best way to debug the variables resulting from the machine configuration? I read about bitbake -e, but the output is a bit confusing. Is there a way to log variables defined inside the machine conf files?
<RP>
this is relatively new, complicated and we do suck at documenting things we know. We do have docs support so a patch will likely encourage them to work out how to improve the docs even if the patch itself isn't perfect first try
<RP>
paulg: not that I'm suggesting your patches aren't perfect :)
<paulg>
timestamp and size of hashserv.db changed too. This might actually work?
<paulg>
The real proof will be if I can go tomorrow (well, later "today") to one of my gutless turds and dodge compiling llvm and mesa
<paulg>
on a similar topic, just how "good" should I expect sstate to be? (hit vs. miss). Say I do "bitbake core-image-minimal ; mv tmp deletemelater ; bitbake core-image-minimal"
<paulg>
should it be able to get to do_rootfs w/o recompiling anything, or is that unreasonable?
<paulg>
or, if I build x86, and then build arm in another build dir (but shared sstate) - should I not see any "foo-native" packages being compiled?
<RP>
paulg: it should skip to do_rootfs in that first case and in the second, it should reuse foo-native
<RP>
paulg: if it didn't and it was the same metadata, that would be a bug
<paulg>
OK, good to know. I'm hoping hashequiv helps. With a daily pull, I typically get depressing results like this...
<paulg>
Sstate summary: Wanted 1249 Local 30 Mirrors 0 Missed 1219 Current 616 (2% match, 34% complete)############################ | ETA: 0:00:01
<paulg>
Sstate summary: Wanted 602 Local 8 Mirrors 0 Missed 594 Current 1254 (1% match, 67% complete)
<paulg>
...in case anyone wondered why I've never had a lot of love for sstate.
<paulg>
I know, most people don't update that often, but I like to see what is going on and what got broke...
<RP>
paulg: we've had a lot of changes to the "core" one way or another. Its is a nightmare for me trying to run test builds too
<paulg>
I don't doubt it one bit.
alessioigor has quit [Quit: alessioigor]
<RP>
Saur_Home: I'm thinking the ton of devtool failures in testing may be from the minicom upgrade. Trying to isolate...
mulk has quit [Ping timeout: 255 seconds]
benkard has joined #yocto
benkard is now known as mulk
<paulg>
[Client ('192.168.168.203', 54928)] Adding taskhash equivalence for c3373afc83602b98ebd45a80b2226a222a8df7fd586453b160a388b20be442d1 with unihash 67b554c8ac220295686ebe70072b98584bfadbf5910c10077679351f19557876
<paulg>
That looks like progress.
<RP>
paulg: yes, its doing "things" :)
<paulg>
...besides heating my house. :-P
* paulg
calls it a win for now and heads off in search of some Zzzz
Thorn_ has joined #yocto
Thorn has quit [Ping timeout: 240 seconds]
Vonter has quit [Ping timeout: 272 seconds]
Vonter has joined #yocto
ptsneves has joined #yocto
ptsneves has quit [Ping timeout: 260 seconds]
simonew has quit [Ping timeout: 256 seconds]
alessioigor has joined #yocto
ptsneves has joined #yocto
ptsneves has quit [Ping timeout: 255 seconds]
jmd has quit [Remote host closed the connection]
jmd has joined #yocto
goliath has joined #yocto
simone has joined #yocto
florian has joined #yocto
bhstalel has joined #yocto
florian has quit [Ping timeout: 272 seconds]
lexano has joined #yocto
bhstalel has quit [Read error: Connection reset by peer]
ptsneves has joined #yocto
risca has joined #yocto
florian has joined #yocto
goliath has quit [Quit: SIGSEGV]
florian has quit [Ping timeout: 240 seconds]
<paulg>
Sstate summary: Wanted 1051 Local 69 Mirrors 0 Missed 982 Current 814 (6% match, 47% complete)
<paulg>
Better than 1 or 2% but still not great. That is after at least 3 machines have built core-image-minimal on master in the last 24h to populate sstate with hashserv, and starting the 4th just 5m ago.
<RP>
paulg: if the config is the same, the second/third/forth should reuse :/
<paulg>
bog standard auto generated config, with only the HASHSERV lines added at the bottom.
<paulg>
It is *possible* that there was/were checkins on master while I was sleeping ; I didn't explicitly verify the SHA; I just pulled before building like I always do.
<paulg>
pulled on one of yesterday's build machines and master didn't change.
<paulg>
would be interesting to see some of those "Sstate summary:" lines from the AB workers...
<paulg>
wondering if their hit/miss ratio is crap too, but they just power through with grunt.
<paulg>
I'm going to run the two tests I talked about yesterday - move tmp aside and rebuild ; new build dir and new arch and build -- all while staying on the same machine.
<paulg>
I have a suspicion that when you move to a physically different builder, something changes the hash. Just a guess at this point.
<RP>
paulg: it would be useful to work out if there is an issue
<paulg>
I've got a mix of ubu-18.04 using buildtools (for newer python) and ubu-20.04 that don't use buildtools. But even then, eventually I should see 90%+
<RP>
paulg: is uninative active?
<RP>
is tmp/sysroots-uninative/ present?
<RP>
(if so, yes)
<paulg>
seems to be - but it looks ancient - dates back to September?
<paulg>
paul@oc-1055t:~/poky$ ls -l build/tmp/sysroots-uninative/
<paulg>
total 20
<paulg>
-rwxr-xr-x 1 paul paul 9045 Sep 1 13:29 relocate_sdk.py
<paulg>
drwxr-xr-x 7 paul paul 4096 Sep 1 13:29 x86_64-linux
<paulg>
-rw-r--r-- 1 paul paul 532 Sep 1 13:27 buildinfo
<paulg>
Let me put it another way - AFAIK, I didn't do anything to actively disable uninative
* paulg
goes to look up what the hell it is (again)
<RP>
paulg: it is enabled by default, I just wanted to check
<RP>
sep is probably fine
<paulg>
checked a bunch of others - also same vintage.
<paulg>
but yeah, I can see how if that special prebuilt libc wasn't there, the sstate would be garbage (how the hell did the name "uninative" come about?!?)
<RP>
paulg: universal native sstate
<paulg>
I'm sure I'll be looking up what it does again in six months...
florian has joined #yocto
<RP>
paulg: for some reason the RTC kernel config patch we just merged comes to mind (goldfish RTC) :)
* RP
thinks he might understand the latest weird race on the autobuilder
risca has quit [Ping timeout: 264 seconds]
<paulg>
another machine that was off-line for a few days. Update and build core-image-minimal
<paulg>
Sstate summary: Wanted 1045 Local 0 Mirrors 0 Missed 1045 Current 820 (0% match, 43% complete)
<paulg>
That is terrible! And so it is building llvm-native AGAIN.
<paulg>
yet I can see that same machine chatting with the hashserv box
<RP>
paulg: compare the filenames in tmp/stamps/x86_64-linux/ see if you can see different hashes
<paulg>
ok
<paulg>
I chose stamps for make-native ; since I'm pretty sure nothing has changed there overnight - ls in the make-native stamp dir and all the hash are different - every one AFAICT.
<RP>
ok, that is a good start as you've a thread to pull on. Is there a make-native sstate artefact in the sstate cache with any of those hashes (do_populate_sysroot I guess)
<RP>
you could also bitbake-diffsigs the two siginfo files, see what it says is different between them. That usually involves chasing up a dependency chain
florian has quit [Ping timeout: 255 seconds]
<paulg>
let me dig around a bit based on the above and quit distracting you for a while. :-P
ptsneves has joined #yocto
<RP>
paulg: quilt-native is always at the end of the dependency chains btw
ptsneves has quit [Ping timeout: 264 seconds]
<paulg>
ok - let me use quilt instead of make.
nerdboy_ has joined #yocto
nerdboy has quit [Ping timeout: 255 seconds]
bhstalel has joined #yocto
goliath has joined #yocto
nerdboy_ is now known as nerdboy
nerdboy has quit [Quit: Leaving]
nerdboy has joined #yocto
kpo has joined #yocto
vladest has quit [Quit: vladest]
vladest has joined #yocto
kpo has quit [Ping timeout: 268 seconds]
luc4 has quit [Quit: Konversation terminated!]
simone has quit [Ping timeout: 256 seconds]
arielmrmx has joined #yocto
florian has joined #yocto
risca has joined #yocto
florian has quit [Ping timeout: 255 seconds]
kpo has joined #yocto
simone has joined #yocto
florian has joined #yocto
arielmrmx has quit [Ping timeout: 240 seconds]
arielmrmx has joined #yocto
ptsneves has joined #yocto
bhstalel has quit [Ping timeout: 264 seconds]
bhstalel has joined #yocto
alessioigor has quit [Quit: alessioigor]
paulg has quit [Ping timeout: 272 seconds]
ptsneves has quit [Ping timeout: 255 seconds]
Vonter has quit [Ping timeout: 268 seconds]
Vonter has joined #yocto
goliath has quit [Quit: SIGSEGV]
simone has quit [Remote host closed the connection]
paulg has joined #yocto
simone has joined #yocto
bhstalel has quit [Ping timeout: 264 seconds]
amitk has quit [Ping timeout: 268 seconds]
paulg has quit [Read error: Connection reset by peer]
paulg has joined #yocto
jbo has quit [Read error: Connection reset by peer]