Segfault during eval
#54285
Comments
|
I think it would help with some more information. |
fatteneder
added
the
needs more info
|
Sadly, this is a private package, so I can't share the code itself, but the error happened when we were evaluating the code with @static if get(VERSION.prerelease, 1, "") != "DEV"
@testitem "JET error analysis TypeOrder package" begin
...
end
endWe're using a patched Julia 1.10.2. |
|
Yeah, but a minimal example is still sought after. If you cannot provide that, then one would need the actual code. Otherwise I believe it is quite hard to know what the actual issue is, no? |
|
Could it be #51267 ? |
|
From the backtrace, it looks to me like a search through module bindings segfaulted, which is different from #51267. I was wondering if #49556 perhaps unintentionally broke something and was hoping the backtrace might be enough to diagnose it. Sadly, this does not happen deterministically, so we have little hope of getting an |
|
It sounds a bit like a call to |
|
We should use the |
…#54671) The race here is that svec might be replaced and a new binding introduced into the keyset while we hold a reference to the old svec, which led to a OOB access on the svec with the index a binding introduced at the same time. This now introduces a bounds check which will force taking the lock if we fail the lookup i.e we had a data race. Fixes #54285 --------- Co-authored-by: Jameson Nash <vtjnash@gmail.com>
…JuliaLang#54671) The race here is that svec might be replaced and a new binding introduced into the keyset while we hold a reference to the old svec, which led to a OOB access on the svec with the index a binding introduced at the same time. This now introduces a bounds check which will force taking the lock if we fail the lookup i.e we had a data race. Fixes JuliaLang#54285 --------- Co-authored-by: Jameson Nash <vtjnash@gmail.com>
…JuliaLang#54671) (#158) The race here is that svec might be replaced and a new binding introduced into the keyset while we hold a reference to the old svec, which led to a OOB access on the svec with the index a binding introduced at the same time. This now introduces a bounds check which will force taking the lock if we fail the lookup i.e we had a data race. Fixes JuliaLang#54285 --------- Co-authored-by: Gabriel Baraldi <baraldigabriel@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com>
…#54671) The race here is that svec might be replaced and a new binding introduced into the keyset while we hold a reference to the old svec, which led to a OOB access on the svec with the index a binding introduced at the same time. This now introduces a bounds check which will force taking the lock if we fail the lookup i.e we had a data race. Fixes #54285 --------- Co-authored-by: Jameson Nash <vtjnash@gmail.com> (cherry picked from commit 20f03dd)
…JuliaLang#54671) (#158) The race here is that svec might be replaced and a new binding introduced into the keyset while we hold a reference to the old svec, which led to a OOB access on the svec with the index a binding introduced at the same time. This now introduces a bounds check which will force taking the lock if we fail the lookup i.e we had a data race. Fixes JuliaLang#54285 --------- Co-authored-by: Gabriel Baraldi <baraldigabriel@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com>
…JuliaLang#54671) (#158) The race here is that svec might be replaced and a new binding introduced into the keyset while we hold a reference to the old svec, which led to a OOB access on the svec with the index a binding introduced at the same time. This now introduces a bounds check which will force taking the lock if we fail the lookup i.e we had a data race. Fixes JuliaLang#54285 --------- Co-authored-by: Gabriel Baraldi <baraldigabriel@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com>
…#54671) The race here is that svec might be replaced and a new binding introduced into the keyset while we hold a reference to the old svec, which led to a OOB access on the svec with the index a binding introduced at the same time. This now introduces a bounds check which will force taking the lock if we fail the lookup i.e we had a data race. Fixes #54285 --------- Co-authored-by: Jameson Nash <vtjnash@gmail.com> (cherry picked from commit 20f03dd)
Hey folks, we saw this segfault in our CI. Any idea what is causing it?
The text was updated successfully, but these errors were encountered: