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
Hi All,
Latest LIT always produces .lit_test_times.txt file. However there is no opportunity to rename it or disable its generation.
This means that when you run several instances of LIT in the same folder of tests, then .lit_test_times.txt can be corrupted. And when it is corrupted, then LIT exits with an error.
Why somebody need to run several instances of LIT?
For example we use farm with lots of machines and each "executor" can run only one thread. So we have to run hundreds of LIT instances to speed up tests execution.
How to fix:
An option can be added that will allow to rename this file.
Something like --lit-test-times-output=FILE_NAME
Add an option that can disable generation of this file.
Generate/read .lit_test_times.txt only when --order=smart. Orders "lexical" and "random" do not use info from this file.
This is how I did it locally.
llvm-project/llvm/utils/lit/lit/main.py
- record_test_times(selected_tests, lit_config)+ from lit.cl_arguments import TestOrder+ enum_order = TestOrder(opts.order)+ if enum_order == TestOrder.SMART:+ record_test_times(selected_tests, lit_config)
The text was updated successfully, but these errors were encountered:
Hi All,
Latest LIT always produces
.lit_test_times.txt
file. However there is no opportunity to rename it or disable its generation.This means that when you run several instances of LIT in the same folder of tests, then
.lit_test_times.txt
can be corrupted. And when it is corrupted, then LIT exits with an error.Why somebody need to run several instances of LIT?
For example we use farm with lots of machines and each "executor" can run only one thread. So we have to run hundreds of LIT instances to speed up tests execution.
How to fix:
Something like
--lit-test-times-output=FILE_NAME
.lit_test_times.txt
only when--order=smart
. Orders "lexical" and "random" do not use info from this file.This is how I did it locally.
The text was updated successfully, but these errors were encountered: