Concurrent call to hs_alloc_scratch() on the same database about hyperscan HOT 2 CLOSED

It is safe to call hs_alloc_scratch() in multiple threads using the same database.

Assuming the calls to hs_alloc_scratch() are successful, and that each thread calling hs_scan() has its own scratch, concurrent use should cause no problems.

Those cases from validScratch() do appear to be the result of using a pointer that isn't pointing a scratch region. Are you able to track the lifetime of those pointers (I'm a fan of printf() debugging), or maybe test using Valgrind?

As an aside: where possible we would usually recommend calling hs_alloc_scratch() once and then use hs_clone_scratch() to make copies of it, as cloning an existing scratch takes less work than allocating and initialising a new scratch, but this is just a performance change and there should be no difference in the end result.

from hyperscan.

The issue was caused by user code, not by Hyperscan. So I am closing it off

from hyperscan.