Release a version including #611 (GC compaction segfault fix) #612

Thanks a lot for merging #611 (74a2232). It fixes a production crash we have been hitting, and we would love to pick it up from RubyGems.

What we see

• tiny_tds 3.4.0 (precompiled x86_64-linux-gnu), Ruby 4.0.7, Rails 8.1, Puma in cluster mode with preload_app!.
• Our before_fork hook calls Process.warmup, which compacts the heap in the Puma master before the workers fork.
• On some deploys, compaction moves TinyTds::Error. Every forked worker then inherits the stale cTinyTdsError pointer and crashes with SIGSEGV (SEGV_MAPERR) on its first connection to Azure SQL:

rb_class_new_instance_kw (object.c:2299)
rb_exc_new (error.c:1473)
rb_tinytds_raise_error
tinytds_msg_handler
_dblib_handle_info_message
tds_process_info
tds_process_login_tokens
tds_connect
tds_connect_and_login
tdsdbopen

• All crashes of a given deploy share the same fault address, across different worker PIDs. Puma restarts the dead workers, and they crash again on their next connection. We had 26 crashes on one deploy and 4 on another. Builds where the class does not move are unaffected, which matches the compaction root cause described in Segfault in connect under multithreaded use: cTinyTdsError/mTinyTds not registered with the Ruby GC #608.

Request

Could you publish a release (3.4.1?) that contains #611, including the precompiled native gems? Until then we will skip heap compaction before forking.

Thanks for maintaining tiny_tds!