RFR: 8294211: Zero: Decode arch-specific error context if possible [v2]

[Aleksey Shipilev]

On Thu, 22 Sep 2022 18:42:24 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

> Good! But why make this conditional with a switch? Who would not want to have better error information?

Because I want to be able to test the generic error handling paths that would run on "generic" arch, without leaving the comfort of my x86_64 machine :)

> src/hotspot/os_cpu/linux_zero/os_linux_zero.cpp line 183:
> 
>> 181:     epc = os::Posix::ucontext_get_pc(uc);
>> 182:     if (ret_sp) *ret_sp = (intptr_t *)os::Linux::ucontext_get_sp(uc);
>> 183:     if (ret_fp) *ret_fp = (intptr_t *)os::Linux::ucontext_get_fp(uc);
> 
> style nits: curly brackets? Remove space in `intptr_t *`?

Right, modified.

> src/hotspot/os_cpu/linux_zero/os_linux_zero.cpp line 390:
> 
>> 388: 
>> 389: void os::print_context(outputStream* st, const void* ucVoid) {
>> 390:   st->print_cr("No context information.");
> 
> Regrettable, maybe something for a future RFE?

Yes. I don't want to copy-paste the whole arch-specific code into this file. The bare minimum to have reasonable `hs_err`-s should do :)

> src/hotspot/os_cpu/linux_zero/os_linux_zero.cpp line 404:
> 
>> 402:   // this at the end, and hope for the best.
>> 403:   address pc = os::Posix::ucontext_get_pc(uc);
>> 404:   print_instructions(st, pc, sizeof(char));
> 
> Does print_instructions not use safefetch like os::print_hex_dump does? If yes, remove comment? If no, should it?

It does use safe fetch after some change in the past. But all other arches also have this comment, so I prefer to be consistent with them. I think this comment is the left-over from the switch to safefetch on this path. That deserves a separate RFR, I think.

-------------

PR: https://git.openjdk.org/jdk/pull/10397