You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Testing this locally reveals that the KILLs are there because ocamlopt infinit loops and starts to take all my RAM and swap space (several GB) because being killed by the system.
Using lldb to investigate seems to indicate some kind of issue in the new Shape module introduced in #10718 (cc @voodoos):
FTR: we've found the source of the issue, and have a fix. But we also tested further on functor heavy projects (irmin) and noticed some compile time degradation due to shapes. We have a couple of patches aimed at improving that.
We're now looking at cleaning them up and producing some benchmarks. We're hoping to open a PR at the end of the week (or beginning of next) with all of that. (And then we can discuss which of those patches we want, if any).
This should be solved with the merge of #10825 . More precisely, the issue with irmin has been resolved and preliminary test on opam shows that irmin was an outlier in term of the complexity of the shape reduction.
When compiling
irmin-pack.2.9.0
with the 4.14 branch (3f170f3), the following problem arise:Testing this locally reveals that the KILLs are there because ocamlopt infinit loops and starts to take all my RAM and swap space (several GB) because being killed by the system.
Using lldb to investigate seems to indicate some kind of issue in the new
Shape
module introduced in #10718 (cc @voodoos):The text was updated successfully, but these errors were encountered: