feat: introduce optional handler strategy by supervacuus · Pull Request #1027 · getsentry/sentry-native

transferring to linux test machine

supervacuus

committed

clean up formatting

supervacuus

committed

fix build

supervacuus

committed

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.

supervacuus

committed

const options param in the handler_strategy getter

supervacuus

committed

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

supervacuus

committed

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

supervacuus

committed

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 nati…

supervacuus

committed

get rid of obsolete local var

supervacuus

committed

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

supervacuus

committed

ensure page-allocator is only referenced on UNIXes

supervacuus

committed

add integration test for managed and native crash

supervacuus

committed

ignore 32-bit Linux build in the integration test

supervacuus

committed

remove native sdk logging assert temporarily

supervacuus

committed

fix pytest skip for x86

supervacuus

committed

inverted demorgan

supervacuus

committed

extract skip_condition to check what provokes the invalid syntax

supervacuus

committed

re-enable log check, since we know it is not about log flushing but dotnet failing to start the handler in CI

supervacuus

committed

clean up sigaltstack initialization

supervacuus

committed

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

supervacuus

committed

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

supervacuus

committed

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

supervacuus

committed

Add changelog.

supervacuus

committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

feat: introduce optional handler strategy#1027

feat: introduce optional handler strategy#1027
supervacuus merged 23 commits into
masterfrom
feat/inproc_handler_strategy

Commits on Nov 15, 2024

transferring to linux test machine

clean up formatting

fix build

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.

const options param in the handler_strategy getter

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 nati…

get rid of obsolete local var

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

remove native sdk logging assert temporarily

fix pytest skip for x86

inverted demorgan

extract skip_condition to check what provokes the invalid syntax

re-enable log check, since we know it is not about log flushing but dotnet failing to start the handler in CI

clean up sigaltstack initialization

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 :-)

Add changelog.