Re: [PATCH] Correct -fdump-go-spec's handling of incomplete types

[Nikhil Benesch <nikhil.benesch@gmail.com>]

On 12/17/20 7:28 AM, Rainer Orth wrote:
> I first tried with the new version included, but that broke badly:
> 
> cc1 -fpreprocessed sysinfo.i -quiet -O -std=gnu99 -fdump-go-spec=tmp-gen-sysinfo.go -o sysinfo.s
> 
> now SEGVs with either infinite or very deep recursion:
> 
> Thread 2 received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 1 (LWP 1)]
> 0x08ea285e in go_format_type (container=container@entry=0xfeffd738, type=
>      <record_type 0xfa50be40>, use_type_name=true, is_func_ok=true,
>      p_art_i=0x0, is_anon_record_or_union=false)
>      at /vol/gcc/src/hg/master/local/gcc/godump.c:688
> 688	{
> (gdb) where
> #0  0x08ea285e in go_format_type (container=container@entry=0xfeffd738,
>      type=<record_type 0xfa50be40>, use_type_name=true, is_func_ok=true,
>      p_art_i=0x0, is_anon_record_or_union=false)
>      at /vol/gcc/src/hg/master/local/gcc/godump.c:688
> #1  0x08ea2cff in go_format_type (container=container@entry=0xfeffd738,
>      type=type@entry=<pointer_type 0xfa50bea0>,
>      use_type_name=use_type_name@entry=true, is_func_ok=false,
>      p_art_i=0xfe6fe174, is_anon_record_or_union=false)
>      at /vol/gcc/src/hg/master/local/gcc/tree.h:3453
> #2  0x08ea3506 in go_format_type (container=container@entry=0xfeffd738,
>      type=type@entry=<record_type 0xfa50be40>,
>      use_type_name=use_type_name@entry=false, is_func_ok=false,
>      p_art_i=0xfe6fe174, is_anon_record_or_union=false)
>      at /vol/gcc/src/hg/master/local/gcc/godump.c:1002
> #3  0x08ea3335 in go_format_type (container=container@entry=0xfeffd738,
>      type=<record_type 0xfa50be40>, use_type_name=<optimized out>,
>      is_func_ok=true, p_art_i=0xfe6fe234, is_anon_record_or_union=false)
>      at /vol/gcc/src/hg/master/local/gcc/godump.c:741
> #4  0x08ea2cff in go_format_type (container=container@entry=0xfeffd738,
>      type=type@entry=<pointer_type 0xfa50bea0>,
>      use_type_name=use_type_name@entry=true, is_func_ok=false,
>      p_art_i=0xfe6fe3b4, is_anon_record_or_union=false)
>      at /vol/gcc/src/hg/master/local/gcc/tree.h:3453
> 
> I've now reverted that part and the build is into make check, as you
> suspected.

Whew, ok, great. Let's leave invalid-godump-2.patch as a curiosity for
Ian, then. The current approach of outputting dummy types for every
invalid type is unsatisfying, but in practice it seems to work alright.

Nikhil