I've been seeing some weird CI failures (https://github.com/bytecodealliance/wasmtime/pull/1263, https://github.com/bytecodealliance/wasmtime/pull/1318, https://github.com/bytecodealliance/wasmtime/pull/1319); @Andy Wortman has a theory that it may disk space related. The error is error: linking with 'cc'
. Can anyone with Azure permissions investigate?
Here's an example: https://github.com/bytecodealliance/wasmtime/pull/1319/checks?check_run_id=506868460#step:16:43
@Alex Crichton, your name came up in the Cranelift meeting on this issue; @fitzgen (he/him) thought you might have insight into this. Also, @Andy Wortman has a separate issue that might also be demonstrating the same behavior
unfortunately I don't really know what's goign on with CI
my best guess is we should roll back nightlies (e.g. use like nightly-YYYY-MM-DD) until one works
I can investigate more tmw when I'm back at work
as a datapoint: building and running wasmtime tests by the github actions actions results in ~17gb of Stuff after running all the steps listed in a nightly build. measured in a fresh clone of the repo, so that's not with any pre-existing target
content.
i notice there are a few steps that are _only_ run for nightly tests, so if it is in fact a disk space issue that might explain why it's only seen on nightly test runs
datapoint 2: building with non-nightly steps results in a build that rounds up to 16gb but is in fact 15.63gb! so we at least cross the 16gb threshold on nightly vs nonnightly
I'm going to try to continue investigating this error today and see if I can get to the bottom of it
ok bisecting a bunch of nightlies at https://github.com/alexcrichton/wasmtime/actions/runs/57523147 and https://github.com/alexcrichton/wasmtime/actions/runs/57538657 shows this is pretty unlikely to be a regression
testing the last good commit -- https://github.com/alexcrichton/wasmtime/runs/514044968?check_suite_focus=true on todays' nightly however shows that this is still failing
gonna try some disk space savings next
ok looks like disk space was the issue -- https://github.com/bytecodealliance/wasmtime/pull/1341, I'll continue to send a few PRs to save disk space on CI so we buy ourselves lots of time before hitting this again
I'm sad the linker didn't print anything more useful, I thought it actually printed out if it ran out of disk space...
that PR saved about 3GB, and https://github.com/bytecodealliance/wasmtime/pull/1343 should save another 7 or so
er, 4*
Last updated: Dec 23 2024 at 13:07 UTC