On 28.06.2019 22:44, Christos Zoulas wrote: > What do you think? SHould we make the code work like before? Or this is > functionality that we don't want to have because it is "dumb"? (I think > that Max here means that it adds complexity and it could be dangerous, > but I am just guessing) Personally I consider it as a mandatory functionality. Maya wrote ATF tests verifying this code path through GDB (64bit debugger, 32bit tracee). On 28.06.2019 22:51, Michał Górny wrote:> The alternative would be to reuse compat32 codepaths more, effectively > switching to 32-bit data types. However, this would mean that debuggers > would have to switch between structures passed to ptrace() based > on whether the tracee is 32- or 64-bit. We want to pass all structs in the format of debugger always (so 64bit in 64bit debugger). Otherwise we would need to do compat32 fixups in userland code in LLDB/GDB. This is done in Linux/GDB and considered as a bug that hurts badly (like siginfo struct fixups). So it means that our previous support was doing the right thing. We still need to do some magic on the debugger side (like picking FS/GS depending on the mode), but that is not a big problem.
Attachment:
signature.asc
Description: OpenPGP digital signature