The old behavior was breaking interop for the PROCESS_BASIC_INFORMATION struct's PEB* PebBaseAddress field. When this field was changed to a PEB[], the struct became a 'managed' type (in interop terms) yet the CLR's interop layer has no information about array length, etc. and marshaling would fail.
It turned out this "small" change led to a series of test failures due to other latent bugs that only manifest when we changed the mixture of how structs (with their managed and unmanaged variants) interact, leading to this surprisingly large set of changes.
The old behavior was breaking interop for the
PROCESS_BASIC_INFORMATIONstruct'sPEB* PebBaseAddressfield. When this field was changed to aPEB[], the struct became a 'managed' type (in interop terms) yet the CLR's interop layer has no information about array length, etc. and marshaling would fail.It turned out this "small" change led to a series of test failures due to other latent bugs that only manifest when we changed the mixture of how structs (with their managed and unmanaged variants) interact, leading to this surprisingly large set of changes.
Fixes #904