- Nov 14, 2024
-
-
-
Georgy Moshkin authored
-
Georgy Moshkin authored
-
Georgy Moshkin authored
-
Georgy Moshkin authored
-
Georgy Moshkin authored
-
Georgy Moshkin authored
-
Georgy Moshkin authored
-
Georgy Moshkin authored
-
Maksim Kaitmazian authored
-
- Nov 13, 2024
-
-
-
-
-
-
-
-
-
Егор Ивков authored
-
Егор Ивков authored
-
Егор Ивков authored
-
-
Егор Ивков authored
-
-
- Nov 12, 2024
-
-
-
Dmitry Rodionov authored
Due to doctest compilation model they've used a lot of memory leading to OOMs in CI.
-
Maksim Kaitmazian authored
-
Dmitry Rodionov authored
-
- Nov 11, 2024
-
-
Dmitry Rodionov authored
Test run and build are now unified with Makefile.
-
Dmitry Rodionov authored
-
-
-
Кирилл Безуглый authored
-
Кирилл Безуглый authored
-
- Nov 08, 2024
-
-
Кирилл Безуглый authored
-
Кирилл Безуглый authored
-
- Nov 07, 2024
-
-
- Nov 06, 2024
-
-
Georgy Moshkin authored
-
-
- Nov 05, 2024
-
-
-
This patch boils down to the following changes: * Register the hook for all cli commands (previously only `run`). * Print the whole text as one log line to prevent splitting. * Always capture backtrace regardless of RUST_BACKTRACE and such. Currently, we act under the assumption that every panic is severe and should abort the whole process. This **does not** have to always be the case -- for instance, we might recover from a mild panic in Sbroad's query planner if the application state hasn't been corrupted. A happy outcome isn't completely ruled out. In the future we might change this behavior by replacing the uncompromising global panic hook with a combo of strategically placed `std::panic::catch_unwind`s and a panic hook for non-unwinding panics. At the moment of writing, `PanicHookInfo::can_unwind` is still unstable.
-