wasmtime-c-api failing with link error on CI? · wasmtime

Stream: wasmtime

Topic: wasmtime-c-api failing with link error on CI?

Chris Fallin (Apr 17 2020 at 17:57):

See https://github.com/bytecodealliance/wasmtime/pull/1536/checks?check_run_id=596044450

CI builds fail with note: collect2: error: ld returned 1 exit status while building wasmtime-c-api. I'm seeing this on several PRs, so I don't think it's related to the changes in question. Is anyone else seeing this?

arm64: Support bool constants by cfallin · Pull Request #1536 · bytecodealliance/wasmtime

Chris Fallin (Apr 17 2020 at 17:58):

Also, I can't repro locally with that command, but I haven't verified my manifest is exactly the same

Dan Gohman (Apr 17 2020 at 18:08):

It's possible the CI is running out of disk space again

Chris Fallin (Apr 17 2020 at 18:10):

Perhaps, since there are no other errors from ld... is there a way to check?

Alex Crichton (Apr 17 2020 at 18:26):

I'll take a look at this

Alex Crichton (Apr 17 2020 at 18:26):

unfortunately last time we hit this we couldn't ever tell

Alex Crichton (Apr 17 2020 at 18:26):

I just repro'd CI locally and saw the target dir was huge

Alex Crichton (Apr 17 2020 at 18:37):

looks like we're running around 8GB

Alex Crichton (Apr 17 2020 at 18:37):

which shouldn't be that large

Dan Gohman (Apr 17 2020 at 18:41):

Does Rust link with the system ld, and is the system ld the BFD ld? If so, it could be running out of memory. How hard would it be to install gold or lld on the CI?

Alex Crichton (Apr 17 2020 at 18:41):

It just shells out to cc

Alex Crichton (Apr 17 2020 at 18:41):

which typically uses the system ld

Alex Crichton (Apr 17 2020 at 18:41):

which is probably bFD

Alex Crichton (Apr 17 2020 at 18:45):

trying https://github.com/bytecodealliance/wasmtime/pull/1543 to get more information about disk sizes

Try to debug directory sizes on CI by alexcrichton · Pull Request #1543 · bytecodealliance/wasmtime

Seeing if we can narrow down failures right now to see if it's OOM or out-of-disk

Alex Crichton (Apr 17 2020 at 18:46):

if we turn off debuginfo the target directory goes from 8GB to 3GB locally

Alex Crichton (Apr 17 2020 at 18:46):

should implement split-dwarf

Alex Crichton (Apr 17 2020 at 18:49):

/dev/sda1 84G 76G 7.9G 91% /

Alex Crichton (Apr 17 2020 at 18:49):

looks like we're out of disk space

Alex Crichton (Apr 17 2020 at 18:49):

man 8gb is pretty small...

Till Schneidereit (Apr 17 2020 at 18:50):

Puzzlingly so: https://help.github.com/en/actions/reference/virtual-environments-for-github-hosted-runners#supported-runners-and-hardware-resources

Alex Crichton (Apr 17 2020 at 18:50):

even 14GB is kinda small...

Till Schneidereit (Apr 17 2020 at 18:51):

yeah

Alex Crichton (Apr 17 2020 at 18:51):

Filesystem      Size   Used  Avail Capacity iused      ifree %iused  Mounted on
/dev/disk1s5   380Gi   10Gi  208Gi     5%  487477 3982052923    0%   /

Alex Crichton (Apr 17 2020 at 18:51):

apparently macos has 200gb available

Till Schneidereit (Apr 17 2020 at 18:52):

wait, how do those numbers add up?

Alex Crichton (Apr 17 2020 at 18:52):

uh...

Alex Crichton (Apr 17 2020 at 18:52):

you're not wrong

Alex Crichton (Apr 17 2020 at 18:52):

Filesystem      Size   Used  Avail Capacity iused      ifree %iused  Mounted on
/dev/disk1s5   380Gi   10Gi  208Gi     5%  487477 3982052923    0%   /
devfs          182Ki  182Ki    0Bi   100%     630          0  100%   /dev
/dev/disk1s1   380Gi  160Gi  208Gi    44% 4296293 3978244107    0%   /System/Volumes/Data
/dev/disk1s4   380Gi  1.0Mi  208Gi     1%       1 3982540399    0%   /private/var/vm
map auto_home    0Bi    0Bi    0Bi   100%       0          0  100%   /System/Volumes/Data/home

Alex Crichton (Apr 17 2020 at 18:53):

~partitions~

Alex Crichton (Apr 17 2020 at 18:53):

(?)

Till Schneidereit (Apr 17 2020 at 18:53):

ok, fair. I didn't know the size column is for all partitions

Till Schneidereit (Apr 17 2020 at 18:54):

/me now wonders what is on Data ...

Alex Crichton (Apr 17 2020 at 19:21):

ok https://github.com/bytecodealliance/wasmtime/pull/1544 is at least a start

Move most wasmtime tests into one test suite by alexcrichton · Pull Request #1544 · bytecodealliance/wasmtime

This commit moves most wasmtime tests into a single test suite which gets compiled into one executable instead of having lots of test executables. The goal here is to reduce disk space on CI, and t...

Alex Crichton (Apr 17 2020 at 19:22):

should give us about 2GB of headroom

Andrew Brown (Apr 17 2020 at 20:40):

#1544 is failing but should we merge it... for the sake of everything else?

Andrew Brown (Apr 17 2020 at 20:40):

e.g. my ConstAddr fix failed on master for space reasons

Alex Crichton (Apr 17 2020 at 20:46):

it's failing though for the same reasons

Alex Crichton (Apr 17 2020 at 20:46):

so it's not actually fixing the issue

Alex Crichton (Apr 17 2020 at 20:47):

still working on that...

Alex Crichton (Apr 17 2020 at 22:22):

ok I think this should be fixed now

Last updated: Nov 22 2024 at 16:03 UTC