* Provide version information in DLL

* Update changelog

* only include rc file when building dll

* move rc code gen, change product name, make FileVersion the same as ProductVersion

* Further clean-ups and reuse for crashpad

* moved code to utility function
* that function uses the target argument to generate all required files (only manual parameter is the file description)
* switched from manual string replacement to `configure_file()` which creates another intermediate file
* to generate the final *.rc file we use file(GENERATE) together with `TARGET_FILE_NAME` generator expression to retrieve "OriginFilename" during the generate-time (between configure and build phase).
* ensure that unit-tests don't pick up unnecessary resource-sources

* update changelog

* update changelog

* replace `TARGET` parameter with `TGT` for good measure

* add test-fixture checks for binaries

* conditionally include windows-specific module

* don't include utils.cmake only for shared-libs build...

...because even if we generate static client libraries crashpad_handler.exe and crashpad_wer.dll should still be versioned.

* update crashpad submodule to getsentry ref

---------

Co-authored-by: Mischan Toosarani-Hausberger <[email protected]>

* add emitting of transaction data inside context trace data

* format

* fix test name

* update test + proper data store

* remove unnecessary test

* remove unnecessary test from tests.inc

* update CHANGELOG.md

* renamed transaction.extra -> transaction.data

* apply scoped transaction/span data to every in-scope event

---------

Co-authored-by: Mischan Toosarani-Hausberger <[email protected]>

* reintroduce #ifdef symmetry regarding the usage of the handler_strategy (which must work on all UNIXes) and the query towards the handler_strategy option which must only work on Linux.

* ensure we "leave" the signal-handler when we invoke the CLR/Mono runtime handler

* ensure the page_allocator is only enabled when we have an actual native crash, we don't allocate before

* continuing the signal chain at the end is something we want for both strategies, because CHAIN_AT_START will reach this execution-path only if the runtime-handler decided that it was an actual native crash.

* add trace logs to at-start chaining, so we can see the behavior in the field when debugging is enabled

* ensure page-allocator is only referenced on UNIXes

* add integration test for managed and native crash

* ignore 32-bit Linux build in the integration test

* extract skip_condition to check what provokes the invalid syntax

* clean up sigaltstack initialization

* overwrite it only when the flag `SS_DISABLED` is set and the query didn't result in an error
* if the query was successful but the flag is anything but `SS_DISABLED`, only log the size and flags of the current stack
* if the query failed then log the corresponding error

* clean up run assertion on output, so we can actually see what's happening.

* create a non-faulty `sigaltstack` for the GHA runner

* disable the test on ASAN runs since that would require an instrumented runtime :-)

dotnet/runtime#13458

* Add changelog.

…1082)

* Add loadNativeLibraries method to allow pre-loading .so files

* Update Changelog

* feat: provide the thread-context to the [on_crash|before_send]-hooks on macOS from the breakpad backend.

* Get rid of remaining static functions with `sentry__` prefix in the breakpad backend.

* update `breakpad` submodule to our fork's `handler` branch